Episode #96 | The Future of Surgical Digital Twin with Gilly Yildirim (CEO, Vent Creativity)

How well do we really understand the body? For decades, surgeons have relied on static scans and flat 2D models to plan procedures. Gilly Yildirim believes it’s time to expand our view to more dimensions. As founder and CEO of Vent Creativity, he is bringing together point clouds, digital twins, and physics-based AI to capture movement with a level of precision that static imaging is far from.

In conversation with Jenny Chen, Gilly Yildirim describes how his team built a platform that sees the body not as a collection of bones, but as a dynamic system of forces and interactions. Their software models the physics of ligaments, cartilage, and bone to predict how a patient’s knee will behave before surgery even begins. The goal is to give surgeons a more clear and reliable picture of what they’re working with.

Yildirim shares the thinking that shaped his company’s approach. The team’s goal wasn’t to design a product, but to build a service that integrates into real clinical practice. Vent’s technology includes Minerva, an adaptive AI engine trained on real anatomical data. It powers Hermes, an FDA-cleared knee planning tool, and inVENT, a cloud platform that lets surgeons explore patient-specific digital twins in full 3D.

Gilly Yildirim has 20 years of experience in orthopedics, so he has unique understanding of the clinical and technical sides of surgical innovation. He discusses how biomechanics, imaging physics, and computational modeling converge within Vent’s framework to create accurate, reproducible results. The same methods used to map a knee could soon extend to hips, shoulders, cardiac systems, and even to full-body digital twins that integrate data across multiple organs and modalities. His work points toward a future where medical planning is not based on snapshots, but on simulations that mirror the patient’s unique physiology.

This episode offers a deep, imaginative look at how AI, physics, and human creativity are coming together to build the next generation of surgical intelligence.

Chapters

• 0:00 Introduction
• 1:03 Vent Creativity's Backstory
• 3:30 Automating Segmentation And Planning
• 5:56 Platforms: Minerva, Hermes, inVENT
• 11:41 From 2D Images To 4D Motion
• 15:46 Validation: Cadavers To Clinics
• 18:11 Data Strategy And Unbiased AI
• 22:39 Building The Team
• 31:50 FDA, Regulation, And Trust
• 37:05 UX, AR, And Surgeon Workflow
• 49:00 Beyond Ortho: The Digital Human
• 51:00 Energy, Compute, And AI’s Cost
• 1:06:36 Final Thoughts & Major Challenges

Transcript

SPEAKER_00: 0:00

Hi there. Welcome to the Lattice Podcast episode number 96. In this episode, I was fortunate to have Gilly Gitterin, founder and CEO of Vent Creativity, to sit down in my studio and have a deeper dive into his company. Vent Creativity aims to create AI-powered to go into autophore circle funding. We'll discuss how Vent Creativity to platforms including data required are maintained to work on a transforming feature for specific data, exploring your workforce, and raising the bar for certificate. Please listen to the disclaimer at the end of this podcast. Hello. Hi. Welcome to the quote unquote studio.

SPEAKER_07: 0:52

That's amazing looking.

SPEAKER_00: 0:53

Thanks for coming, Gilly.

SPEAKER_07: 0:55

Thanks for having me.

SPEAKER_00: 0:56

And I want to introduce really quickly, Gilly is the CEO and co-founder of Ven Creativity. Is that right?

SPEAKER_07: 1:01

That's right. Founder.

SPEAKER_00: 1:02

And how did you come up with this company? Why did you tell us a little bit about what it does and you know your inspiration, et cetera?

SPEAKER_08: 1:09

Yeah, I've been in uh orthopedics and medical devices for about 20 years now. And really the issue was I have started in robotics as industrial level and then moved that over to biomechanics and everything else. That's how I got into orthopedics. Then I worked at industrial robots for uh robotics and surgical applications. Then uh worked on additive manufacturing, as you know, uh 3D printing implants, and uh worked on biologics and various different things. Really at the end of the day, even though we're building all these technologies, there were incremental upgrades on existing technologies.

SPEAKER_07: 1:45

So there was really no change in healthcare and outcomes for patients. And our goal was to create systems and surgical plans that were patient-specific. And that really wasn't available until we had AI and tools that could really speed things up.

SPEAKER_04: 1:59

Yep.

SPEAKER_08: 2:00

Uh so just timing felt right right in the middle of COVID. We had nothing better to do. I found some engineers and said, all right, we're gonna automate all these things so that all these pain points surgeons face. How do I get the STICOM file to create some meaningful thing without having to have six million people working for me? And then how do I create reports? And then how do I get better outcomes? Uh, we decided we can fill all those gaps and understand how to solve them. And that's how the company really started. Initially, we started with really understanding the surgical plan, and then we found gaps of segmentation. It was just costly and not very automatic. So we create our own. And you know, slow by slow we found different tools that we really needed and then added those on as tools that matter to us.

SPEAKER_07: 2:47

Having done that, the company became really more consolidated because now it's a standalone product. It doesn't really require outside help to function, uh, which means it helps customers. They don't have to shop around for six different products, they can work with one product and it's all cloud-based, so it's easy to access. And it reports automatic, so they don't have to fiddle around figuring how to write reports out of that.

SPEAKER_08: 3:09

So, you know, at the end of it, it's like a really boring product to solve problems driven by very um sophisticated AI.

SPEAKER_07: 3:17

But you know, surgeons don't like to see technology thrown in their faces. So our goal is to really reduce the amount of complexity to them and then give them real insights, you know, five or six insights they can really take action on that's gonna really make a difference.

SPEAKER_00: 3:30

Yeah. You know, I have a theory. Sure. A theory is you have to be in the orthopedic industry or device industry for 20 years to make things work.

SPEAKER_02: 3:39

Right.

SPEAKER_00: 3:40

Because um deep industrial knowledge and relationships really matter in the space. Um and the other thing that surprises me is that you you founded this company in 2020, right? I assume in the middle of the pandemic. And back then, I actually remember there were quite a few companies claim they have automated, automated segmentation tools for either 3D printing or surgical planning or yeah, virtual surgical planning.

SPEAKER_02: 4:06

Yeah.

SPEAKER_00: 4:06

And I'm just surprised to hear that yet another company needed to be found to actually do that.

SPEAKER_08: 4:12

Yeah, they still claim they have it, and uh, I don't think most of them do. They do in a certain extent. So it really comes down to data and uh what you do with it. Right. So a lot of the companies that have auto-segmentation are usually segmenting really pristine bones, pristine objects.

SPEAKER_03: 4:29

Right, right.

SPEAKER_08: 4:30

And uh that's fine, it depends on the application. For sports medicine, it makes sense. But when it comes to arthritis, the bone is not what you expect.

SPEAKER_00: 4:38

Right, imperfect bones.

SPEAKER_08: 4:39

So imperfect bones. And uh on top of that, we're not really just creating shells of the bones. It's not like uh here's an outline and do something with it.

SPEAKER_03: 4:48

Right.

SPEAKER_08: 4:49

So even out of magic manufacturing, as you know, sealing the whole object is difficult. When it's a bone, it's even more difficult because there are holes and things in the bone where capillary is going. So it makes a mess when you're trying to print that thing.

SPEAKER_04: 5:04

Right.

SPEAKER_08: 5:04

So our system doesn't really use meshes, meshes that are byproduct of point clouds that we're using.

SPEAKER_04: 5:10

Yeah.

SPEAKER_08: 5:10

So these point clouds are essentially the entire data from the scans, CT, MRI, fluore x-ray, doesn't matter. And uh that gives us the full picture. Then we can isolate different regions for different purposes, whether that's the outline of the bone or density regions for various um decision making. So, you know, if we go back to 3D printing, we could print multiple different phases of the bone. So you could have a very realistic object that has different stiffnesses. So if you want to cut it, it would feel realistic to the surgeon because you know there will be a cortical bone, cancellus bone, and they can really interact with it. Uh, we actually had some pants on that to do remote surgery. So, you know, a surgeon can cut their bone in a kitchen and then the robot can do it, you know, 5,000 miles away in the real patient. So there's a lot of cool applications that we haven't really touched yet.

SPEAKER_00: 5:57

Yeah. Um, you have three core platforms. Is that correct? I mean, I just got this information off the website. You want to just unpack a little bit, especially for the one that congratulations on getting the FDA clearance, which is the Herma's knee platform. And then you also have some something called Miner Minerva and Invent. Would you like to just unpack a little bit what they do briefly?

SPEAKER_08: 6:19

Uh yeah, we like to label things, it's fun.

SPEAKER_00: 6:22

I know. I'm like, how did you come up with Hermes? Um is your wife involved in this?

SPEAKER_08: 6:27

Yeah. Uh she's definitely in the marketing. And um typically what I do is I'm trying to understand. Well, I'm from Turkey, so there's a lot of Turkish Greek um uh gods and things that go into it because it's fun.

SPEAKER_04: 6:40

Yeah.

SPEAKER_08: 6:40

Uh Minerva is sort of uh random offshoot, I guess that's Roman. But uh Minerva is the platform technology. Uh it's like the system that doesn't really know anything but does everything. So coming back from my robotics background, PLCs, programming logic controllers don't know what they're doing, but you can program to do certain sequences of things. Right. So Minerva doesn't really know what it's doing, but it has a rules-based approach. It's a pure AI that says, I'm gonna try to keep this human upright or achieve some goals, whatever the goals are, minimum energy, et cetera. And then Minerva can be applied to different things, one of them being Hermes. And Hermes is specifically for knee. And then there'll be a hip application, the application, um, shoulder application, et cetera. But essentially the idea in a Hermes is it takes um rules from Minerva, such as segmentation, landmarking, and surgical planning. Okay, and applies it to a very specific application of the knee.

SPEAKER_04: 7:34

Okay, got it.

SPEAKER_08: 7:35

And uh, Invent is our basically data center. Um online UI that you can access. Either you look at the plans, everything looks great. It's all pre-planned for you, so you accept it. Or you can make changes or you can customize things. So the we have digital twins, what we call them. So digital objects that you can introduce into the point clouds, so then you can see how that interacts with the bone. So, for example, a cylinder could represent a drill hole or a screw, or uh a block could represent a cut, or a sphere could represent a sphere, um, like a reaming operation.

SPEAKER_00: 8:08

Yeah.

SPEAKER_08: 8:09

So you can do those virtually and then see what the bone looks like around that.

SPEAKER_00: 8:13

That's the invent, right? I just want to clarify. Yeah. I think I saw one of the YouTube videos by the Venn Creativity YouTube. And I will share that link with everybody. Now, you you mentioned a really interesting concept, which is digital twins. We know suddenly five years ago, everybody in 3D printing industry started talking about digital twins and this giant marketing campaign out there. That's right. While it is pretty fascinating, and also we can totally just conceptually understand what that means. Can you just tell us what are some of the real challenges you have faced to actually create a product that can you can actually call yourself digital twin, of course?

SPEAKER_08: 8:52

Yeah, I think the digital twin is uh a really difficult area, and there isn't a lot of players out there. So it really started with um land surveying, uh, where um you can have geo um measurements of the surface and then do certain things like plotting for buildings, etc. Yeah, and more recently construction where you know where pipes are, etc., by looking through augmented reality to know where to cut, where not to cut, to make sure you don't um you know go through a pipe, et cetera.

SPEAKER_03: 9:19

Right.

SPEAKER_08: 9:20

So we really are actually using the same technology in terms of point clouds. It's all coming from surveying and self-driving cars. So the advantage we have is with point clouds, it's easier to interact with objects than a mesh object, which is really difficult.

SPEAKER_01: 9:34

Right.

SPEAKER_08: 9:34

Because um, you know, engineers out there would know whenever you're trying to do a fine element model, it's difficult. And after hours and after competition, it might just crash and you start over and you don't know why it crashed. Uh with point clouds, uh, because it's individual points, you can interact with them and see point-to-point interactions easily. And uh let's say a bone is one object, ligaments and muscles are other objects. So then we can create objects and start assigning materials part of it, and then have them interact with each other with collision detections. So physics models borrow from our uh partners in NVIDIA. We're part of NVIDIA and inception programs, so we have live resulting to us. So then we can move. Um, essentially, what happens is right now the gold standard in the world is 2D imaging, and you take a CT or X-ray, you look at it in 2D slices and make a decision.

SPEAKER_04: 10:27

Yeah, that's what I do every day.

SPEAKER_08: 10:28

Right. Which is um very funny because humans are used to 3D worlds, but in radiology and surgery, everybody's looking at 2D images to make a decision. Even in um top uh robotic systems, you create a three-dimensional model and then slice it to 2D so that people can look at it in 2D where the implant's gonna fit. Yeah. Um, but in our case, we take 2D slices, create 3D point clouds, and then we animate them in 4D. So now we know how a human knee is gonna move, for example, because we know how the ligaments are gonna interact with the bone.

SPEAKER_04: 11:00

Yeah.

SPEAKER_08: 11:00

So without us, so Minerva figures out how this bone is gonna move around without us telling him here's how it should move, because our goal is to remove all of the bias from the system. So we don't want to bias it in any way that's going to change the way it's gonna decide. We say, Here are the ligaments available to you, minimize the energy and tell us how that thing's gonna move. Then introduce an implant and say, all right, now try to fix that motion so it's more stable. And that's sort of Minerva basically planning a surgery without knowing what it's doing, and that's the most stable solution that's possible.

SPEAKER_00: 11:34

Yeah. So it totally makes sense. That said, um, it's not easy.

SPEAKER_06: 11:40

No.

SPEAKER_00: 11:40

Um, one question I just because since I'm not from the engineering background and you're certainly a lot more knowledgeable than me, the concept of pixelated 3D printing, is that similar?

SPEAKER_08: 11:52

So yeah, I think um it's sort of our cousins in terms of whenever anybody has any image, right? The first thing we do is create a 2D image, and that 2D image is actually uh either an AI or manually uh manipulated image. So we're taking signal, softening it so it's not blocky. And then in three dimensions, those are called voxels. So the voxels are essentially sort of a cube representation of the point. Right. So now essentially, uh not to get too technical into it, but essentially if you have a cube instead of actual shapes, obviously you're gonna be off by a certain amount. Right. Yeah. So we're removing that bias by saying here are the points that are available to us, and we're at the resolution of the image. It doesn't matter, whatever that is.

SPEAKER_00: 12:37

Yeah. I would have guessed that's a huge amount of data you have to process.

SPEAKER_08: 12:40

Right. So it's 20, 30 million points per bone. So we're talking maybe 100 million points per scan. Uh but the amount of information you get from that, especially when you colorize it, is amazing because things that people don't think are possible. So cartilage, meniscus, and ligaments from CT, or um, you know, bone density or bone models from MRI, etc. So it's very sort of cross-functional. Once we create a point cloud, we can do a lot of things with it in our own system. And that really explains how a human body evolves. So we're calling an evolutionary alignment. How do you evolve to sort of stand up straight versus the primates? And we can figure out where the implant should go based on that evolutionary alignment. So that removes all the biases of here's a point area, so that's a landmark.

SPEAKER_04: 13:30

Yeah.

SPEAKER_08: 13:30

Because that point area, if you look at you know, data, certain reserves are off by almost two centimeters in any direction. So that's not a very reliable landmark to begin with. We remove those landmarks and we just say where do ligaments and muscles attach and how do they attach? And landmarks don't really matter to us because how the ligaments move matter to us.

SPEAKER_00: 13:51

Right. You added a lot more dynamic information into what used to be usually static. And honestly, when I was looking, you know, at your company, I was thinking, why can't just regular diagnostic radiology do the same thing? Because I read MRI lombard spine every day. But you know, when we get MRI, we lie in a scanner. It's a supine position. So your nerve root and your spinal cord and your discs are all in your supine when you're lying down. People are hurting when they're standing up and walking around. You know, same thing as a knee arthritis, for example, right? How the abrasion of the damaged cartilage is causing, you know, rubbing against each other, how that's causing ligamentous imbalance and stuff like that. We could have done that potentially, isn't it?

SPEAKER_08: 14:38

Yeah. Interesting thing is, um, you know, there's this thing called wolf law. Wherever the bone is needed, it grows. So what happens in arthritis or even healthy knees is that you can see where the density is because that's where the patient spends most of their time. It's sort of a snapshot of their life. Right.

SPEAKER_00: 14:56

It's a static one point in life. Yeah.

SPEAKER_08: 14:58

Even if they're laying down, we already know how those two bones come together normally when they stand most of the time.

SPEAKER_04: 15:04

Yeah.

SPEAKER_08: 15:04

So then we can figure out what their natural alignment is instead of uh forcing some sort of a position for an implant or reattaching an ACL. So we know the exact tightness we need to achieve that position. And if the position is wrong, then we know sort of globally average, as well as phenotyping, where that um density patch should be. And we've done some studies in the past with implants where we show that the bone grows back to where it's needed if the implant is positioned properly. So the implant is forcing it to sort of bone to grow back to normal locations. So we can do both things, um, pre-operative or postoperative analysis of where the bone density is and to see if we achieved what we're trying to set out to do.

SPEAKER_00: 15:45

And then do you have any kind of validation to validate what you're simulating and predicting so far?

SPEAKER_08: 15:52

Yeah, so uh we just started cadaver studies uh with Orlando Health Partnerships. Uh we have our surgeons there. And what we've been doing almost every two weeks is we're doing cadaver studies where we're removing the cadaver bones um with their soft tissue intact. And the surgeon moves them passively as well as loading it in different directions to figure out the soft tissue envelope. And then we're checking our uh digital twins against that to see how far off we are. And the goal is not to sort of be able to guess up front what it is. Our uh physics simulator is gonna be very fast, and the way it's very fast is we're training it with data from those labs. So it's gonna learn over time on the properties that it's really expecting, and then it gets better and better over time. And then the next step will be clinical studies where obviously the muscles and ligaments are active in a patient.

SPEAKER_01: 16:41

Yeah.

SPEAKER_08: 16:41

And even in surgery, they're not active. So then, you know, we have partnerships with a company called orthopedic-driven imaging, and uh, they have a fluorosystem. So we're able to see how the patients move actively in real world and then feed that back to our system to see how that differs from a cadaver or a patient who's under anesthesia. So it's um, you know, people always say, Oh, you know, how do you know? Um, we don't, you know, we don't pretend to know everything. We're sort of building incrementally.

SPEAKER_03: 17:09

Right.

SPEAKER_08: 17:09

But the gold standard right now is essentially 2D images. Right.

SPEAKER_00: 17:13

It's is it's amazing how powerful it can be, actually.

SPEAKER_08: 17:16

I think we're creating improvements, but we're not you know pretending that we're finished everything, we're sort of building towards more value added. But anecdotal evidence, our FDA cleared products and uh quite a few surgeries now, and the outcomes have been very good, positive outcomes from uh surgeon's perspective. And uh, we'll start formal protocols for clinical studies to see how we're doing on outcome scores that are accepted.

SPEAKER_04: 17:40

Yeah.

SPEAKER_08: 17:40

Uh obviously they're very subjective. So we're looking to again uh fibroscopic imaging to see how the kinematics changes in terms of patient mobility, flexion angles, etc., yeah, that are more predictive of sort of outcomes than pain scores and others that are very subjective.

unknown: 17:56

Yeah.

SPEAKER_08: 17:56

So that'll be the next steps for the company.

SPEAKER_00: 17:58

I mean, AI in medicine for this type of medicine um is expensive. Um people think it's you just have to say AI, then the magic genie would just give you what you want.

SPEAKER_06: 18:10

Right.

SPEAKER_00: 18:11

One, you have a data challenge. How are you gonna acquire the kind of data that's actually useful for your simulation? Because everybody uh get a different scan, slightly different technique, dosage of the CT, MRI. I don't know if you use MRI data yet. Um how do you manage to acquire this kind of and this amount of data? Um, what's your data strategy, in other words? And and and maybe you can tie in like how it's a secret.

SPEAKER_08: 18:40

Then everyone helps.

SPEAKER_00: 18:42

If it's a secret, don't tell us. Um and how does you know these various um AI partnership, for example, the NVIDIA inception program that you just mentioned help you in that regard?

SPEAKER_08: 18:53

Yeah. Yeah, so um I think coming from a background in orthopedics, as you said, 20 years. Yeah. Uh I forgot to sort of uh chime in on that. Uh you know, I would say I cheat because I you know 20 years of things that didn't work, so it's easy to know what needs to be done to fix it. And having worked with all the surgeons and radiologists over the 20 years, yeah, they all know uh my thought process and they all understand what I'm doing. So they all believe in the product. So they've been very upfront and very open to partnerships. So we're working with a lot of radiologists and surgeons who sign partnerships to share data.

SPEAKER_04: 19:33

Yeah.

SPEAKER_08: 19:34

And uh we're using that data, uh, you know, basically allow them to do use it for research in compensation. And we're doing a lot of research-based publications as well as data sharing to understand how it can improve their outcomes for their surgical grades systems. So the goal here is um, you know, upfront, uh, this actually we'll talk about the 510k process later, but our goal has always been completely the identified data because, as I said, uh we're obsessed with uh unbiased systems.

SPEAKER_04: 20:05

Yeah.

SPEAKER_08: 20:06

So I didn't want to know anything about the patients, from their gender to their health to their age to whatever. And FDA wanted to show that we're, you know, represent the entire demographic of the US. We're like, we don't know because we scrub all that data. So we actually have to go back to our surgeons and get information directly from them to see uh, you know, which scans we actually ended up getting. Uh but ironically, our goal has um being unbiased, which means that all the data we're getting is getting broken down to um statistical methods. So going back to my phenotype um answer, um, there's phenotyping called CPAC. Um we're not believers in that. We don't think it works properly. It's completely biased, but we use three-dimensional phenotyping again using point clouds. So we have you know hundreds of thousands of points that we can phenotype a patient with. It's very similar to sort of shape morphing, but in the other direction. So it's already shaped. We're trying to figure out groups, and we're using um you know Gaussian models to figure out the phenotypes. And our system is not supervised or an unsupervised AI, it's both. So we always create an initial condition that says we think the number is 18 phenotypes, then let's nudge it. We nudge it plus or minus five types and see how many people are jumping groups. So in anything we do, we don't sort of settle for one way or another. We're always sort of challenging the system. So now it's popularized as agentic systems, but we're sort of created our system to be agentic where we have internal systems that are fighting each other for um, you know, who's right. And oftentimes when we have one way of doing something, we usually have two or three other things that are also trying to do the same thing to see if they all agree on the same solution from very different perspectives, whether that's bone density or ligament lengths or stiffness or uh multiple other things, have to decide on where the plane of the cut should be compared to a biased mechanical axis, which is sort of a plumb-drop line. So everything we do builds into the AI structures that are the core of the company, because if you don't have a clearly defined AI structure, then you're sort of you know grasping at straws to figure out how to solve the next problem. But we have a clearly defined sort of culture of the company, but it's an AI culture of the company where we know how to solve problems based on how we go about um you know step-by-step solutions.

SPEAKER_00: 22:38

Yeah, that's impressive. I mean, thank you. I read an article recently talking about robotics, since you're from also the robotics industry, that one of the major challenges in robotics is the software side of things. It's expensive to actually build things from ground up. Yeah. But yet I think you're actually quite frugal with uh your capital allocation right now because I know you guys are still early stage, right, in terms of funding. Um so you you founded the company in 2020, which is also not an easy time to actually get a team together. You want to just tell us a little bit of that journey? Because I feel like I missed out on that.

SPEAKER_08: 23:14

Yeah, um, I would say it was the easiest part of the company, uh, because you know, most engineers were sitting at home with no basics. Nothing to do. Exactly. So uh if you leave an engineer for a few days, they're gonna get bored and they're gonna want to do something or solve something.

SPEAKER_03: 23:28

Right.

SPEAKER_08: 23:29

So um I'm from Columbia University, biomedical engineering, and now uh MBA as well. Uh so I went back to Colombia and advertised, and I said, you know, I'm looking for engineers and uh what Columbia is known for is medical imaging and AI. Uh so radiology and AI and BME.

SPEAKER_04: 23:47

Yeah.

SPEAKER_08: 23:47

So it's perfect. So I was like, you know, I'll take two. So my first employees uh were actually from there, and uh they were just out of master's program, so they just started their careers with me. So it was sort of you know getting them up to speed on a company startup culture while at the same time trying to build a minimum viable product to show. And um we really had a very interesting model. So we spent a lot of time creating a very fast prototype model because our goal was not to create a product, our goal was to create a service first. So it really bootstrap. And the goal here was working with large companies and say, you probably have a problem with X, Y, Z. We can help you with that. Uh so they want to find out, going back to the example, is the implant gonna grow the bone back to where it needs to be.

SPEAKER_04: 24:36

Right. So their RD pipeline, basically.

SPEAKER_08: 24:40

So they would pay um, you know, a small amount, very large in our eyes, very small in their eyes. Yeah. So we solve their problems. And the irony is, I mean, obviously that gets us through a few months, uh, no problem. But the irony is large companies never want to touch any kind of product development, co-development agreements. They just want to pay for research and then that's it. And then as they grow and grow, they're like, wait a minute, maybe we should have um had a product development agreement. So yeah, it's sort of that's how you know you grow companies growing when all of a sudden those research agreements are not that easy to come by because all of a sudden they have a kind of internal algorithm of when they're gonna their eyes are gonna lit up. So yeah. That that you can feel the tilt in the company. So now we're definitely a growth stage company because everybody wants to work with us, but maybe not for just sort of paper play, but something more. But yeah, when we first started, um it was just a few engineers, and the goal was really solving problems fast so that we can build our core products at the expense of doing research for others. And that's just sort of um model that I thought made sense because uh being a deep deep tech company, nobody was going to really understand what we're doing until we could show them something. And uh my joke is you know, like Bane from Batman, nobody cared about what I was doing until I put on these ligament models. As soon as we had some ligament models, everyone was like, oh my god, this is amazing. But before that, nobody could understand what we're working on. So that took five years to sort of get to that level. But until then, it was just sort of all hand wavy, like here's how it's gonna work, here's what's gonna work. But it was very difficult to sell people on it.

SPEAKER_00: 26:20

So you started with uh getting some bootstrap money uh by consulting for larger companies. And uh when this these platforms are uh I'm assuming they're generating some kind of revenue or are they appre-revenues still?

SPEAKER_08: 26:35

Uh right now they're sort of a bit of both. So we're doing sort of research grade models, yeah. And then our FDA clear product, we're still negotiating on insurance. Um there's a code for it. But you know, uh hospitals have to approve that. So that's sort of where we are. Make sure that we have a case in point with a couple of our hospitals. Once that's proven out, then every other hospital is comfortable with it. That's always difficult because there's not a clear hardware that can easily be coded. It's more sort of, you know, gray areas like, okay, so how does this add value?

SPEAKER_04: 27:10

Yeah.

SPEAKER_08: 27:10

And that's at the word of the surgeon, so it's sort of chicken and egg. We have to get over that hump.

SPEAKER_00: 27:16

Yeah, I have to say, you know, in the last couple months, believe it or not, um, just on my dashboard for pitch 3D pitches, yeah, there are more and more um digital virtual surgical planning. Yeah, oftentimes AI, AI driven. Um what do you think the the future of the space is gonna be? I mean, what was your what was your initial vision for Vent? And then after five years, now today, has that evolved over time?

SPEAKER_08: 27:47

Yeah, absolutely. So I think initially I was five years ahead of everyone on sort of auto planning for surgery, but that was never gonna be a differentiator because everybody could auto plan at some points. The AI is available to achieve that. So I think we we basically noticed that, and then we had some differentiated and point clouds and everything else. So it really went from pure um surgical planning, which anybody does with a bone mesh model uh with some landmarks and whatever, to really uh understanding the bone's own structures and sort of how to plan it. And then now in the past year or so, all the soft tissue and ligaments, which would be very difficult for everyone else to do without our core technologies.

SPEAKER_04: 28:33

Yes.

SPEAKER_08: 28:34

So we really went from surgical planning to a full digital twin of the anatomy to see how it works and how it could work better, which is uh when you think about it, it's not where do I cut to fill this gap. It's more how do I restructure this entire anatomy to stand up? So it's a very different, it's complete opposite direction of coming from it than uh how do I place an implant to how do I fix this human?

SPEAKER_00: 29:01

It's right from the adding all the other complex factors into physiological informations and bone density, for example.

SPEAKER_08: 29:09

It's a big mode, I think, because it takes a lot of undoing to start over for other companies to go to that direction.

SPEAKER_00: 29:16

Um how about finite element simulation and that's sort of more uh also kind of functional evaluation of those structures. Are you guys incorporating that or is it something in the in the books?

SPEAKER_08: 29:31

Or we have some things in the works, but again, um we're not looking at it as an engineering or scientific tool to use in surgery. It's more of a uh educational tool for surgeons. So our goal is not to have very uh strict and perfect um solutions to say here's the exact stress with one misses, whatever. Yeah, it's more uh there's a is there a fracture risk or not if we place this thing here.

SPEAKER_03: 29:59

Mm hmm.

SPEAKER_08: 30:00

So that uh again going back to our physics models, there's new solutions that can create fast and actionable items. Uh the reason for that is any FEA in a bone structure that is very complicated will take days or weeks.

SPEAKER_00: 30:17

Yes.

SPEAKER_08: 30:17

So not very useful.

SPEAKER_00: 30:19

Even with India's latest trips.

SPEAKER_08: 30:21

So I was just in a conference in Rome. Yeah. There were a lot of amazing papers in biomechanics, but you know, admittedly, there were like this took weeks, and n equals one. You know, it'd be great if we could scale this to clinical grade. I'm like, it would, but it's not possible.

SPEAKER_04: 30:36

Yeah.

SPEAKER_08: 30:37

So there's some shortcuts that we figured out where we think it's going to add quality. And, you know, the question's always going to be, you know, how do you know it's going to work? Right. But the answer is, you know, what's the penalty? So like if if it's an acceptable location for implants that's going to minimize the risk of the fracture in our minds, would you go for the higher fracture risk because visually it looks like it might fracture? Or would you go for the solution just because it's sort of so it's it's more of a trust building issue than uh I think explaining physics to surgeons because they're not going to read uh finite elements literature and say, yes, it was published and peer-reviewed, so it's perfect. Uh we might do that, but we think that our proprietary systems would benefit better if we didn't open that up to everything. So our goal is to just say uh here's a physics model, take it or leave it. Right. You don't have to use the factor models, you don't have to use the attachment models for bone fixation, etc. But uh I think there's a lot of burden of truth that now that really stifles innovation. And we would like to um sort of give advice and sort of make it on take it or leave it to kind of an advice on the side, especially for fine on that level.

SPEAKER_00: 31:50

Yeah. I mean, maybe maybe we don't need to make a relatively simple problem more complicated, and especially because we have technological limitations at the moment. I mean, we just don't have enough computing power. The going back to that article I was referring to about robotics world, I mean, same thing. Is there a lot of things that are good to have? I I would like to have someone to brew tea for me as robot, but that would take trillion dollars to just get that thing right. Um excellent. Now, I mentioned earlier during those podcasts that you achieved FDA queries for Ermay's knee platform this year. Yeah. Um but there are a lot of stories behind it, and also you have a history in regulatory science.

SPEAKER_07: 32:38

I do.

SPEAKER_00: 32:39

And how did you end up there? What a random journey.

SPEAKER_08: 32:42

Yeah, uh about 30% of the company is regulatory and quality people.

SPEAKER_00: 32:46

Wow.

SPEAKER_08: 32:47

So we're saying we're a very boring regulatory company that happens to have some AI in it. Um because, you know, we're in sort of it is a regulative space. Yeah, we're in Bay Area right now. I think uh it's sort of a reaction to the Bay Area. There's a lot of software development here that is about getting to the market fast. And there's a lot of stories of large and small companies getting into healthcare and getting out really fast without an explanation. Yeah, not to throw too much shade, but uh really uh my engineering background uh always go back to it because engineers go to school for four years or maybe more, and then they get out and make things. And I'm always confused because you know I think the core of it, what engineering is defined is as defining requirements and then risks, and then solving the risks to minimize them to meet the requirements. But then I think everybody just forgets that on the door and says, all right, let's just build stuff, and then if it breaks, we'll just fix it. Put some duct tape on it. Uh and our goal has always been um let's build regulatory and quality structures first, yeah, and then build um tools and then software based on those tools, and then go to the market with it with the FDA's approval because I think uh FDA is now more and more stringent on AI.

SPEAKER_00: 34:08

Oh, really?

SPEAKER_08: 34:09

We know this.

SPEAKER_00: 34:09

Um because this is the first uh first time I heard about those, actually.

SPEAKER_08: 34:13

Yeah, so we're actually the I think the sixth company in orthopedics to get FDA clearance for AI products.

SPEAKER_00: 34:19

Uh-huh.

SPEAKER_08: 34:19

And overall, I think maybe in the order of a few hundred, if not less. Wow. And it kind of in the older days, I think AI was sort of nebulous to them. But in the last two, three years, because there was such an avalanche of AI tools, especially the language model side of things, yeah, they locked down and said anything that uses any kind of neural networks or machine learning is AI. It doesn't matter if it's sort of checking your name or if it's you know segmenting bones. So uh maybe two, three years before we could have passed easily. Right.

SPEAKER_00: 34:55

I thought it was like six months to clearance.

SPEAKER_08: 34:57

Typically it's three months for a decision, and then usually you have questions. So six months probably about right.

SPEAKER_04: 35:04

Yeah.

SPEAKER_08: 35:05

But um, it took us much longer because we had to update our tests and show more um evidence. And that's that's a good thing, I think, because um while we did a lot of work on our segmentation and landmarking, sort of all those things that are AI-based, right? Uh they were validated by students, and they have to check each time they use it. So it's not sort of here it is, cut here. They have to approve each step and say segmentation looks correct, landmark looks correct, cut planes looks correct. So there's a lot of check stuff. Uh but I think FDA is going to be overwhelmed with all these AI tools coming in, not just in language model, but also in segmentation and bone modeling, etc. So I think it makes sense to look into these and say, you know, are these really meeting the customer needs? And they're not just sort of doing something that is maybe not necessarily dangerous, but it's also not meaningless. Oh, yeah.

SPEAKER_00: 36:06

I mean, there are a lot of meaningless uh software, AI software in radiology, for example.

SPEAKER_08: 36:12

I keep hearing we're gonna get value-added, uh value-based care in the US. Yeah, yes. You know, I'm holding my breath and I'm about to suffocate. But uh I would love to have that because I'd love to prove that our system is adding value to the system. And um, you know, some of the older technologies, they came, but I don't think they added too much value. So um, but you know, having said that, I think they paved the way for technology for people to understand technology matters in the OR and in the healthcare. So it was great they did able to do that, but the the burden of proof is very low. I'd like it to be higher so that we can say we meet that burden.

SPEAKER_04: 36:52

Yeah.

SPEAKER_08: 36:53

Because if you're not increasing patient outcomes by at least 5-10%, so then what are you doing? You're just adding another technology for the surgeons to have to deal with.

SPEAKER_00: 37:02

Right. I mean, that's actually one of my questions.

SPEAKER_08: 37:04

Right.

SPEAKER_00: 37:05

Because like I said, have so many companies coming through us just pitch to the small platform along in terms of automated pre-surgical plenty of various body components. I mean, if I'm a general surgeon who, you know, raw hospital have to take care of basically everything, right? And I have all these tools, how am I gonna know which one actually is worth my time to learn and implement and pay for?

SPEAKER_08: 37:31

Right. So there's um, yeah, there's a conflict between first and second mover advantage in healthcare.

SPEAKER_00: 37:37

Yeah.

SPEAKER_08: 37:38

I can never tell which one is better.

SPEAKER_00: 37:40

But I think you're a first or second.

SPEAKER_08: 37:43

I don't know. It it switches from day to day. Uh first mover, in terms of uh, I think our field is very human, even though it's software and AI.

SPEAKER_04: 37:53

Yeah.

SPEAKER_08: 37:53

Because the surgeon really has to trust me.

SPEAKER_04: 37:56

Yes.

SPEAKER_08: 37:57

That me, as the person who created this, knows what they're doing. So then by default, what I created should be correct as well. Because at the end of the day, I think there's gonna be a lot of distrust in AI because it's a black box. Um, how do we know this thing is actually giving us the answers? Half the time I have to check to see why was that solution, because that's an interesting solution. Uh, I had comments like, oh, that was a different sizing plant and different location that I would have normally put. I tried it and it was very stable. So, you know, sometimes it doesn't make a lot of sense in terms of conventional solutions.

SPEAKER_04: 38:29

Yeah.

SPEAKER_08: 38:30

So it's a very human solution that they have to trust me and say, okay, you know, they didn't write all this time.

SPEAKER_04: 38:35

Yeah.

SPEAKER_08: 38:35

They're gonna be right again. There's a lot of burden on that too, right? Yeah, because I have to be right every time. Otherwise, your first move advantage goes away.

SPEAKER_04: 38:43

Yeah.

SPEAKER_08: 38:44

Second motor advantage, I think I'm less worried about in my field because uh in 3D printing, there's always a better, faster, or different technology that comes. Uh, and somebody can say, I'm gonna spend another few thousand dollars to get another printer because this one is better for XYZ reasons.

SPEAKER_00: 39:01

Yep, every day. There's something new.

SPEAKER_08: 39:04

Um, I think going back to your comments on surgical planners, I think that's second move advantage.

SPEAKER_03: 39:09

Right.

SPEAKER_08: 39:10

Because both surgeons and radiologists hate software for one reason or another. If you know how to solve their pain points, then second mover just looks at the problems and then solves them. Great. Uh, but that's I think for sort of incremental technologies.

SPEAKER_04: 39:25

Yeah.

SPEAKER_08: 39:26

If we're going to ground them up, it's very difficult to follow. It means you have to start from scratch and believe that we're doing something right. And I think if there's second movers in my fields that are working directly with me, that's validating that I'm doing the right thing. So I'm happy to welcome them. So please uh by all means enter digital twins.

SPEAKER_00: 39:47

Yeah, well, a couple of comments. First of all, I I do agree, I think relationship building with the clinicians is a barrier to entry to scale up your product. And uh I have my personal, I have personal experiences with new technology, pack system, EHR systems, and the customer service part from the software provider makes a huge difference of whether or not we're gonna continue to use something.

SPEAKER_02: 40:13

Yeah.

SPEAKER_00: 40:14

And you know, when certain when service degrades or going away, that's when we actually stop purchasing. So yeah, building those relationships are huge. And the other thing I want to mention when you said that there's a huge um, not huge, but a resistance to new software and stuff like that. Recently, I just realized my mother is using Chat GTP. She's in her 70s and she enjoys it. She's creating all sorts of stuff every day. And I'm just like, you know what? She's not supposed to know how to use this. Why is she using this? And my conclusion is because they made the UI UX so simple that a 70-year-old who didn't know how to use computers don't understand engineering, can totally just play with it.

SPEAKER_02: 41:02

Yeah.

SPEAKER_00: 41:02

So 3D 3D printing, for example, 3D printer, I got a couple of 3D printers and nothing ever worked. And I'm fairly skilled. And I think the bottom line is the technology just hasn't, the UI UI except the technology hasn't evolved to the point where it can scale sometimes is the problem. So, what do you think of developing a platform that's friendly to people who are gonna use it?

SPEAKER_08: 41:30

Yeah, a couple comments on that. I think the barrier to entry is very difficult. But um, you know, in Chat GPT and other models, if there's no barrier to entry, it's free. Then you're gonna try it.

SPEAKER_01: 41:41

Yeah.

SPEAKER_08: 41:42

And you're gonna have an experience. Uh I personally love and hate it.

SPEAKER_01: 41:46

Me too.

SPEAKER_08: 41:47

It never works on code, it just drives me crazy. And it works great in you know, text that I need to sort of adjust for whatever reason. So, you know, I use it with a grain of salt every time. And um, the interesting thing is um the access to entry, I think, is always, you know, I'm an engineer, so I'm guilty as charged. I'm creating engineering products, hoping that surgeons will understand it. So up until this point, maybe four years into the company, we built things that we knew were gonna be good because I know what needs to be fixed. And um I worked at a large company for seven years, working directly with consultant surgeons. So I knew exactly what they were asking, and I knew exactly what we would give them, and not necessarily what they were asking, because oftentimes they're looking for the faster horse.

SPEAKER_01: 42:40

Right.

SPEAKER_08: 42:40

And I know I can do much better than that. So good point. Then creativity has always been, yeah, yeah, yeah, we know what you want. You know, a 3D planner for your mechanical access, but that's not gonna solve anything. That's gonna be the same technology, automate it, getting the same outcomes, useless. So we sort of listen, but at the same time, we do what we know is gonna help. But at the tail end, now in the past year, the priority has been user experience. So understanding how to take uh a PAX or a radiology system and getting rid of six million buttons to say, here are the five buttons, they're all very obvious. And if you click this, this is gonna happen. And if you click here, this is what happens. So uh the goal is almost to not have user instructions. We have to have it because FDA mandates it, but you should not need it. It should be obvious to you. And you know, when you click something, what's gonna happen? Very much like Apple.

SPEAKER_00: 43:38

That would be amazing because I'll never read.

SPEAKER_08: 43:40

Right. I'm an engineer, I don't read instructions, my wife. But um, the goal here is um translate to 2D images now, and then also 3D in augmented reality. Because my biggest pet peeve is augmented reality is heavily used for showing really large screens, which is ridiculous to me. And I think I'm worried that AR is gonna go away again. It comes and goes every 10 years. I I know it's about to collapse again because a lot of the companies are falling down and now it's turning to consumer glasses, which is not very useful. The issue is I think lack of imagination on what to do with AR. It's always the same. Uh we can show instructions, we can show directions or whatever. So in surgery, what we're doing is we're not showing giant screens of radiology images for you to act on, but we're displaying three-dimensional bone models and ligaments and cut planes and where to place an instrument virtually so you can sort of move your retractor there, etc. So it's real meaningful tools that you can use and improve your surgery. But we're a decidedly software company. So if the hardware goes away, we'll have to go to navigation and robotics, and there will be no AR. So the worry is always I wake up in the morning and Google loves stalking me. So it tells me what's going on in the world of AR or AI every morning. So I don't really have to do too much research. And I open YouTube and it gives me the latest and greatest on the things that I need to know. But yeah.

SPEAKER_00: 45:16

That's that's a very interesting perspective because usually, you know, most of the people I interview are from the hardware side. They don't worry about the hardware going away because that they represent that. Right. And you're the software provider. Now you're worried about the conduit of your software, is may experience a downturn.

SPEAKER_02: 45:35

Right.

SPEAKER_00: 45:35

I think Meta is still gonna be there. I think Apple is gonna be there though.

SPEAKER_08: 45:40

They are, but um, obviously they're very consumer-focused companies. True. So what they're creating is for consumers, and a lot of the glasses I need have um, you know, trackers and inputs that are not available in consumer level products. I see. I can always you know pivot and figure out how to use those, but uh I think they're not necessarily gonna have the burden of proof that FDA requires for medical grade. So, you know, for better or worse, we'll see how it shakes out probably in a year or two. Hopefully, there are new glasses that are industrial grade coming out.

SPEAKER_04: 46:14

Yeah.

SPEAKER_08: 46:14

Uh Magic Leap is sort of what we're working on right now, and they're partnering with Google. So the hope is that they're gonna have more advanced classes out. Uh looks like Google's creating our own consumer level glasses.

SPEAKER_00: 46:26

Yes, with Warby Parker, I think. Yeah.

SPEAKER_08: 46:29

Assuming they're using Magic Leap fans and technologies uh to augment it. But hopefully there's an industrial level uh progress because uh I think there's a lot of applications for it.

SPEAKER_00: 46:41

There are ways, Billy. There are ways. Where are you in Silicon Valley? You just have to go down to the coffee shop down the down the street and and hang out. That's what you need to do. Now, have you actually scrubbed into operation with your software ever?

SPEAKER_08: 46:56

Uh with my software, uh actually no. Ironically, no. I've been very uh remote. Uh I've scrubbed into hundreds of surgeries and cadaver labs.

SPEAKER_04: 47:06

Okay.

SPEAKER_08: 47:07

But um, it's been um so a lot of the places that are currently using my technology are away from me. Uh this is my uh wife's job. I'm actually going back and forth. Our office is in New York.

SPEAKER_00: 47:19

So um so where are your clinics? Where who are using your software currently located?

SPEAKER_08: 47:23

So right now we're active in Orlando Health, okay in Orlando.

SPEAKER_00: 47:27

Right.

SPEAKER_08: 47:27

And then HSS in New York.

SPEAKER_04: 47:29

Okay, yeah, I know about that.

SPEAKER_08: 47:30

Right. Uh that's a good hospital. Yeah. And uh Anderson Clinic has used it several times. And now hopefully we're talking to some Stafford surgeons, and so we'll get that started.

SPEAKER_04: 47:40

Yeah.

SPEAKER_08: 47:40

And uh we're talking to uh some other major clinics uh in the United States. Hopefully, we'll announce them soon. But the goal is uh it's what's called a limited market release and healthcare.

SPEAKER_04: 47:50

Yeah.

SPEAKER_08: 47:51

Uh five or six sites just to get the growth pains out of the way in terms of data structures and dealing with PACs, uh, if anybody has to deal with it. So how do you transfer data is always a difficult thing.

SPEAKER_00: 48:03

Right. The the tiniest step that we never anticipated would have a problem. Right. Now, have you are you planning to go to OR at some point? Because I know other medical device people would.

SPEAKER_08: 48:14

Yeah, yeah. So um I think um I'd like to go to every OR, and the reason is we picked handpicked these ORs because they're varying in sizes and uh volume in terms of how many surgeries they do, and also the implant types.

SPEAKER_03: 48:32

I three.

SPEAKER_08: 48:32

So the goal of Vencreativity is to be agnostic to everything as well as unbiased. So it should work on any medical imaging at any resolution. So we're shooting for the lowest common denominator and resolution, yeah. As well as any implants and any surgeon's hands in any hospital. That way the goal is sort of rising tides. So I'll have a very busy schedule once we're up and running, just going to dig in the case. I know, I can't imagine.

SPEAKER_00: 48:58

Yeah, but it's impressive. Your goals are very ambitious, um, but I hope it became a reality. So I can't wait to hear your update next. Now, in terms of the clinicians you kind of got in touch or currently working with, not to be ageist, but uh is there a difference talking to different generations of surgeons?

SPEAKER_08: 49:20

Yeah, I mean, uh I think I mean I I usually pick surgeons who are open to technology. I have to, because it's gonna be a very difficult conversation. Uh, but yeah, I think I should mention we're probably one of the only technology companies that addresses sort of 100% of the market for orthopedics. Because um, when you're talking about navigation and robotics and augmented reality, that's really representing anywhere between 20 and 30 percent of the markets uh because they have to be open and have the capital upfront expense to build and buy those. Right. Versus 70% of the market is still looking at an x-ray and sizing an implants based on literal transparencies from you know 80s uh high schools and saying looks about right, and then they have manual instruments to place in and then cut.

SPEAKER_04: 50:08

Yeah.

SPEAKER_08: 50:08

So our system is open to x-rays and other imaging. So we can say instead of your usual three degrees that you always do, maybe do two or whatever. So you know, with augmented reality, we can refine a manual surgery to any small practice as well as manual surgeries. So the goal here is to expand 100%, and um that only happens when we can go back, you know, going back to trust issues.

SPEAKER_01: 50:36

Yeah.

SPEAKER_08: 50:37

Uh older and younger surgeons all have to trust the system. And it's obviously more difficult for older surgeons because not because they're older, but because they've done so many surgeries, they're like, they don't need I've done this thousand times, I don't need your help.

SPEAKER_01: 50:50

Exactly. Yeah.

SPEAKER_08: 50:51

So, you know, my job is more about uh negotiations and trust building than uh selling a technology. And I think uh that's where I excel in terms of as an engineer. I'm not sort of like a full engineer, I'm like a I used to, what was I called myself? Communicator, essentially, a bridge between an engineer and a surgeon. So I understand what both sides are saying, and somebody has to translate in the middle. So I know what they're gonna respond to. They're gonna respond to you know better whatever flexion axis. They're gonna respond to, you know, less time in the or they're gonna respond to less complications, better outcomes, you know, uh more surgeries because you know most of them are getting paid less and less per day. So, you know, you really have to know your audience. And having talked to a lot of these advisors in my previous roles, I know more or less what they respond to. And you know, this also goes with VCs and every other group. Uh, you never go into any communication in my field and say, here's my pitch or here's my saletaker later. You have to be very fast on your feet and watch their eyes and watch their emotions and say it's not sticking, pivot.

SPEAKER_04: 52:07

Yeah, yeah.

SPEAKER_08: 52:08

And it has to be so fast that it doesn't feel uh defensive. So it's a whole thing you have to build as a as a founder, and it's not something you're born with. So you just have to try a few hundred times.

SPEAKER_00: 52:20

Personalized, personalized pitch for everybody. Exactly.

SPEAKER_08: 52:23

Personalized surgery and personalized pitches.

SPEAKER_00: 52:26

That's a good salesperson.

SPEAKER_08: 52:27

Exactly.

SPEAKER_00: 52:27

You know, there are actually papers published recently about 3D printed pre-surgical models or some sort of similar along that line. And they got feedback from different levels of surgeons who are, you know, surgeons who are very new in training versus those who are highly experienced. And usually the highly experienced people just it doesn't improve their outcome, doesn't improve their economics. And on the other hand, I mean I would say economics, I would walk back a little bit. I think it will improve their economics, except they haven't thought about it a lot of times.

SPEAKER_08: 53:04

I have a love and hate relationship with personalized um cutting guides. Yeah. Because you're sort of stuck with what was planned.

SPEAKER_01: 53:11

Right.

SPEAKER_08: 53:11

There's no way around it at that point. So either 100% ditch it or 100% go for it. So I think back in early 2000s, they were working on um plastic guys that were flexible so they could sort of change alignment on the fly if needed.

SPEAKER_00: 53:27

Yes, yes, I heard actually they still do. They still do, they can actually adjust the cutting guide now interactively.

SPEAKER_08: 53:33

It needs to happen because then you're stuck. You're sort of then the surgeon is frustrated because they have to throw it away and then go to plan B.

SPEAKER_04: 53:40

Yeah.

SPEAKER_08: 53:40

But uh I think really uh what it comes down to with um 3D printing and guides is um, yeah, again, trust building with the surgeon. They have to absolutely know what they planned and they have to know how it's gonna fit, and they have to be happy with the plans, otherwise, it just turns into frustration.

SPEAKER_00: 53:59

Yeah. Well, it's funny that you talk about surgical guide, because that's not a topic I was planning to talk about, but I have a lot of background in that. Yes, and also the field is definitely getting disrupted by, I mean, just the surgical guide alone itself is getting disrupted by the robotic side of things, all the orthopedic implants companies now have their own various robots to work with, their implants. And you guys, yeah, the virtual surgical planning. Now, are you available interoperatively? I'm assuming, or is this only purely pre-operative?

SPEAKER_08: 54:32

Right now, clear for pre-op, and we are about to design freeze our interop.

SPEAKER_00: 54:36

Okay.

SPEAKER_08: 54:37

So in interop, we have we're going to sort of probably have two um applications at the same time. One will be your typical infrared tracking with typical cameras. And then the second will be the augmented reality tracking, the infrared.

SPEAKER_04: 54:51

Okay.

SPEAKER_08: 54:52

And we're also working on direct bone tracking, so there will be no need for trackers at all. So that's sort of that's the gold standard is you put in your glasses, it shows you where the cuplanes are without any trackers, and you start cutting. So there's zero uh friction going into surgery.

SPEAKER_00: 55:08

And this is superimposed with all your uh wave point or cloud-based um system.

SPEAKER_08: 55:13

Yeah, you can turn things on and off with your voice so your hands are free. And you can say internal ligaments.

SPEAKER_00: 55:19

The density map, the dynamics of I mean the dynamic information on the ligaments as well, interoperatively.

SPEAKER_08: 55:25

Yeah, that's the I think our value add is you can see all the soft tissue and ligaments. So if you're putting tension and you're trying to see if that's gonna be a viable location for an implant, or if you put like uh what's called a um trials. Yeah. So they're not quite implants, they're just plastic pieces. You can move the knee around and then you can see virtually if the ligaments are getting strained or not. Sometimes one ligament gets strained and there's just no way around it. But the surrogate sees through their glasses exactly which fiber to release. That could be like a tiny nick, and that's gonna make a whole difference. So the patient doesn't have pain. So we can give instant feedback, and they can also check intro-op if the pre-op plan was correct or not. If it's not, you can ask for a recalculation based on the ligament tensions. So all of that is without a physical tensioner, which is typically a hardware um created system that's you know, adding all, you know, going back to value base, adding more hardware and more things to work with for the surgeon.

SPEAKER_04: 56:23

Yeah.

SPEAKER_08: 56:24

So the goal here is in the intro version, um, either augmented reality or just typical robotics showing where to cut. And uh going back to your comment on patient-specific guides, we don't have patient-specific cutting guides. We have an interesting solution, I think, is we can 3D print uh an angel wing, which is essentially um something you put on a cuts uh guide. So your typical off-the-shelf cut guide you would use for manual surgery.

SPEAKER_04: 56:52

Okay.

SPEAKER_08: 56:53

There's a cuts and slots that the stall would go into. You can put our little angel wing in there and it's instrumented, so you can track it in space. Now your cutting guide is instrumented. I see. So now your um essentially typical manual instruments are became a million-dollar robots. And now you know where to place it exactly as you planned, pin it, and that's a robotic surgery for a few thousand dollars or just a million dollars.

SPEAKER_00: 57:20

You know, I think I I saw a very similar publication on this years ago by a Chinese researcher.

SPEAKER_08: 57:27

Yeah.

SPEAKER_00: 57:27

So yeah, the concept was definitely out there.

SPEAKER_08: 57:30

The concept, uh, I think where we added value in 3D printing is my lack of trust in 3D printers. So you can 3D print a plastic piece and it doesn't have to be accurate. So uh obviously coming from regulation quality, my issue is always okay, now we have to register this thing to make sure it was printed correctly. Right. But I don't care about that. I put the instrument on it, track instrument. Then you can digitize the bottom surface, even if it's completely off from what you planned as a CAD model, then that plane is now accurate compared to the tracker.

SPEAKER_00: 58:04

Do you have a picture of those? So maybe later on I can get it from you.

SPEAKER_08: 58:07

Yeah, I think yeah, I think we should have to do that.

SPEAKER_00: 58:08

So just have a concept of what it looks like. I mean, if you're allowed to share.

SPEAKER_08: 58:12

I mean, yeah, it's it's it's an incremental technology as far as I'm concerned, because our goal is to be trackerless anyway.

SPEAKER_04: 58:19

Yeah.

SPEAKER_08: 58:20

So it's not something we're likely gonna commercialize. Uh we may or may not will see it.

SPEAKER_04: 58:25

Yeah.

SPEAKER_08: 58:25

It kind of depends on sort of, as I said, how the technology progresses.

SPEAKER_00: 58:29

Yeah, I would love to see the interoperative state phase of your product. That now, how do you feel like you feel the competition from the larger players are gonna threaten your survival? Because, you know, obviously J and J, Stryker, they're all working on their own thing internally.

SPEAKER_08: 58:47

Um, no.

unknown: 58:49

Okay.

SPEAKER_08: 58:50

I think uh, I mean, having come from large companies, right? Um, it's I look at it as a life cycle. So engineers like me work there for a while, but on very specific um projects that are adding value.

SPEAKER_04: 59:03

Yeah.

SPEAKER_08: 59:03

And large companies make their money on um implants. So robots sells for a million, let's say.

SPEAKER_00: 59:12

Yes, as a million dollar attract last time.

SPEAKER_08: 59:15

Right. Let's say you sell a hundred robots over the year, and then another hundred, and then eighty, and you know, there are only so many robots you can sell. Right. And maybe you sell software packages, et cetera. But at the end of the day, the software packages and everything else is a cost center for large companies. So they're always looking for new innovation than they can buy instead of build, because internally those engineers are better used for other things than being a cost center. So the sort of industry self releases in terms of innovation. Uh, there's no value in companies building themselves because I'm giving so many secrets away here. But RD versus MA are.

SPEAKER_00: 1:00:00

Yeah. The audio recording is still recording in progress. Okay. Where do we start again? Or are you talking about a large company? Go back to the MA and R. Good stuff.

SPEAKER_08: 1:00:36

I'll start an R D and MA and they'll be like a nice um break.

SPEAKER_00: 1:00:40

Okay. One, two, three. Start.

SPEAKER_08: 1:00:43

Right. So the R D and MA are two different buckets. That's sort of the not so um sudden, uh, not so subtle difference. Uh RD, the goal is to build products that sell at minimum cost possible. MA is buying products. And my role has been pricing out MA compared to RD. So how much can we build it for? How much can we buy it for? Buy usually wins. So the goal here is finding innovative companies, buy them, and roll them into the system. Then it's already pre-built and ready for the markets. Oftentimes already in the market, like us. And um, you know, it costs a lot of money and time to go into smaller centers. Even in the big centers, it takes money and time to enter with software. So I don't see large companies as competitors, I see them as partners. Yeah. So we're all talking to basically all the large players right now in the market because they're all interested in what we're building. And they know that they're not going to build that uh typical life cycle for any kind of medical devices, about seven years in large companies, but and you know, about a year for us or less.

SPEAKER_00: 1:01:54

I have to say, Gilly, this is a secret sauce for success. It's just enjoy the pain point of the larger companies. Um, it's really hard to know unless you're already an industry insider.

SPEAKER_08: 1:02:08

Um it creates innovation. So I think it's it's an interesting cycle where people leave because they know they have a good idea, but they're not gonna be able to build it in a big company. So then it spurs out innovation outside, and then oftentimes they're brought back. So it goes in cycles.

SPEAKER_00: 1:02:27

Yeah, I definitely have a known stars along that line with big uh big implant providers. Right. Now, I think we have really had a great conversation so far. Um, I want to just kind of ending in on a sector of future outlook. Sure. Um, we mentioned there were some industry cycles, and we could be either up or down with various keywords. Um what do you think in three to five years, what do you what do you want to see?

SPEAKER_08: 1:03:02

In healthcare as a whole?

SPEAKER_00: 1:03:04

It's an open end a question in healthcare, in particular orthopedic surgery world.

SPEAKER_08: 1:03:10

Sure.

SPEAKER_00: 1:03:11

And also with Vent creativity.

SPEAKER_08: 1:03:13

Okay. Yeah. I mean, as a as Vents, uh, I think our goal is much larger in orthopedics. Uh, we're a digital twin company in healthcare, yeah. Not in orthopedics. So our goal is to expand a digital human uh to the entire field.

SPEAKER_04: 1:03:28

Yeah.

SPEAKER_08: 1:03:29

Uh our goal is to be the Amazon of healthcare where you can come in and use our marketplace for your specific needs. So um going back to how we were founded in 2020, um, because orthopedic was considered um non-priority surgery, it was shut down.

SPEAKER_04: 1:03:46

Yes.

SPEAKER_08: 1:03:47

So we actually started off with lung analysis, heart analysis, et cetera, to understand, you know, where to find nodules for COVID, et cetera.

SPEAKER_01: 1:03:56

Yeah.

SPEAKER_08: 1:03:56

And uh we're never not forgetting that. So we're right now working on hernia as well as looking into hearts and oncology in other areas. The goal is to see where else we can be useful. Yeah. Looking into a full body scan of MRIs for preventive care.

SPEAKER_04: 1:04:14

Yeah.

SPEAKER_08: 1:04:14

So I think three to five years, we would have made a significant change in orthopedics in terms of outcomes for the patients. We're talking 10-15% increase. The rest of it we probably can't address from you know patient biases in terms of pain, etc. But I think mechanically we can address that with our system. And then uh at the same time, growing other verticals where we can use Minerva for other applications where someone like me is not gonna be able to run that show. We're gonna need people who are experts in those buckets. But I think um I have the vision to sort of start that and bring people that are uh in my mindset to say, how can we break this down because it's not working? And how can we build it back up? So I think digital twinning is gonna expand in uh healthcare with increasing compute power.

SPEAKER_04: 1:05:08

Yeah.

SPEAKER_08: 1:05:09

But at the same time, cynically, I'm not um you know blind to the fact that compute power is drawing a lot of energy. So uh three to five years, we need solutions in energy. And I'm hopeful always for fusion and fission. Um, so I think those are very viable ways of solving a lot of problems. Today I was looking at um geothermal energy, that was very interesting.

SPEAKER_04: 1:05:32

Yeah.

SPEAKER_08: 1:05:32

Uh blasting through the earth to get to that. But energy and water are gonna be major issues for everyone.

SPEAKER_04: 1:05:39

Yeah.

SPEAKER_08: 1:05:40

Uh, but uh at the same time, healthcare is not gonna really progress, or basically nothing's gonna progress if AI is priority, because that's sort of a zero-sum game in terms of the pie. Uh, we can't all compete for energy, somebody has to lose. So hopefully we can expand it with new technologies.

SPEAKER_00: 1:05:59

Yeah, I mean, there are things that we know we don't know, and there are things we don't know we don't know. What I mentioned is that, you know, uh, what is that? Deep seek from China. For example, it's kind of like out of blue kind of scenario. Right. I mean, I don't know the whole story about it, but the scenario of a better foundation model uh could dramatically reduce the kind the kind of energy that we use as a one scenario that it's a possibility. We just don't know if it's there.

SPEAKER_08: 1:06:28

Yeah. Um stand up contrarian, and we need a lot of contrarians who say there's better ways to solve that.

SPEAKER_00: 1:06:36

Yeah. Um now, final reflections here. Um, well, actually, a couple of things. One is um, what do you what is your your major challenge? What are your major challenges right now that you're facing?

SPEAKER_08: 1:06:49

I think major challenge is explain the story. Uh it's always difficult. I think we're still at that phase of people not quite understanding uh what that means, what digital twin means.

SPEAKER_04: 1:07:01

Yeah.

SPEAKER_08: 1:07:02

Um I'm trying to be very careful not to say simulation because I think there's a very difficult difference between simulation and digital twinning.

SPEAKER_00: 1:07:08

Okay, what's the difference?

SPEAKER_08: 1:07:09

Okay. Simulation is essentially understanding more or less how a system moves based on average rules and based on average computing. I think digital twinning is essentially taking the exact human or whatever object and using that exact information to solve it. There's a large difference in terms of uh specificity, but uh there may or may not be a difference in the solution. But if there is, then digital twin should win because it's very specific to that person in the healthcare aspect. I think in Digital Twin, we have to really concentrate on that aspect. And uh again, going back to the whole bias, I think going into the solution, people really need to be very mindful of bias in AI. And I don't see a lot of that. So that's that's my biggest pet peeve in the field is I think you need to go in there with major rules and going back to engineering again. Has to be requirements and rules put in place on how do we reduce bias going in so that it's not disadvantaging any person or any patient or you know, any situation. Otherwise, uh it's an echo chamber. Then it's gonna work for enough people that it's good enough, but maybe that's a marketing tool, not really an engineering solution.

SPEAKER_00: 1:08:29

So maybe maybe your tools can be used for uh for China.

SPEAKER_08: 1:08:34

I mean, I tools can be used around the globe, so we have no boundaries. We are used in Switzerland and Turkey right now, in outside of the United States. And uh we're actually just starting in Belgium uh for research.

SPEAKER_00: 1:08:47

Uh we'll get into the NZRs eventually, but maybe one of these days also to Asia country, because you know it's totally bought different body size and type and bone density.

SPEAKER_08: 1:08:56

Japan is a very large market in terms of technology and orthopedics. Going to Japan tomorrow. Yeah. Because Japanese uh patients are very different than uh Caucasians. Yeah. And the bone structures are not really allowing those implants to fit correctly. Right. So how do we fit to their specific alignments, which is very various? This is gonna probably lose half the crowd, but uh you know, bow-legged essentially. And how do you solve for that so that they're not forced to be a different alignment? They're not, yeah, is essentially our core. So we can solve it, but with implants available to us. And what we found is there's a market for patient-specific implants, only 10%, but that's still 10% that's underserved. So, you know, that goes to patient satisfaction. Yeah. Uh 10% we can solve for mechanical alignment. 10% we need to solve with patient-specific implants and instruments, etc., that are sort of outliers that just can't be resolved with your off-the-shelf products.

SPEAKER_00: 1:09:56

Yeah, I think the 10% is actually probably overestimates. Probably I think the majority of people could benefit just virtual planning. Personalized virtual planning.

SPEAKER_08: 1:10:06

Yeah, planning, yeah. I mean, I meant implants.

SPEAKER_00: 1:10:07

Yeah, implants are difficult because there's a lot of economic issues with and we'll experience the same problem that you mentioned about surgical guides. Yeah. Okay, you can create this expensive personalized implant, but on the day of operation, the patient changed.

SPEAKER_08: 1:10:22

Yep.

SPEAKER_00: 1:10:23

What are you gonna do?

SPEAKER_08: 1:10:24

Yeah, yeah. So yeah, I think uh going back to our fluoroscopy partners, uh, that change. You can have instantaneous surgery down the road where they get a flora, plan is created in 10 minutes, then you go into OR, and then outcome of the OR can be decided when they're getting discharged to see how they're doing.

SPEAKER_04: 1:10:43

Yeah.

SPEAKER_08: 1:10:43

So there's a world where the ecosystem is completely a chain where you don't need to have any delays in the system.

SPEAKER_04: 1:10:50

Yeah.

SPEAKER_00: 1:10:51

Okay, well, final question, I promise, sure is do you have any suggestions for the next generation of entrepreneurs, engineers, stewards?

SPEAKER_08: 1:11:00

Stay in school kids, man.

SPEAKER_00: 1:11:02

Don't do drugs.

SPEAKER_08: 1:11:04

Exactly. Um, I think it's funny because when I was in school, uh, what was the coolest thing when I was in school? It's almost changes every you're always off by 10 years, right? So I think you probably should not go for what's popular right now. So, you know, AI and software right now is obviously the biggest thing. But what's next? When I was in school, I think it was materials because everybody was getting into custom materials and um fibers and you know, nanoparticles, et cetera.

SPEAKER_00: 1:11:34

That is so unsexy right now. I just want to say that.

SPEAKER_08: 1:11:36

No one cares. And then three, it was we were just talking about it earlier. Ten years ago, I was in 3D printing and it was the hottest field ever. I don't think I heard 3D printing in a while. So, not to throw shade at the whole crowd here, obviously, but uh, you know, it's it's gonna come back, but it goes in waves. And then augmented reality and virtual reality, I think it's gonna go over again and then be back up. So it's a matter of, I think, having that internal gut feeling of what's next. And probably not listening to people because oftentimes you're gonna say, well, no one's gonna need that. Um, so I went into this field uh being a hardware and robotics engineer. I pivoted to software. Not because I had this master plan, but I thought, you know, that's how we solve everything. And then AI happens right in the middle of my um progress.

SPEAKER_00: 1:12:28

So that was almost like you picked up something earlier than everybody else.

SPEAKER_08: 1:12:33

Yeah, my you know, I had a few sayings, one of them is fortune favors you prepared. So I think what I would advise to all the students and everyone else is study game theory. Uh I'm biased because I come from an industrial engineering background, but yeah, game theory and operations research is the core of everything you do in the world, really. Very boring level, but essentially not assuming something's gonna happen. Uh my wife calls me pessimistic, but I'm not really pessimistic. I'm sort of prepared because if if I know what's the worst thing can happen, I have a prepared solution for that. Yeah. Versus, you know, hoping it's gonna be the best case scenario. So now I'm prepared for all the different solutions possible. And then hopefully I have a smart contouring part uh who's going to sort of behave uh as you would expect, even if they don't, you sort of have an idea of what's gonna happen. And I think uh engineering principles and these um game theory principles really go a long way as core parts of your life, as well as social skills, obviously. But with these, I think you can't really do wrong because if you're believing a product, then you know how to position it based on you know how people are responding to it.

SPEAKER_00: 1:13:51

Absolutely. You know what? I'm so inspired today. I'm gonna read these right after this podcast. Well, thank you so much for coming over and I really enjoyed this conversation. And hopefully we can have another catch-up thought sometime in the future.

SPEAKER_08: 1:14:06

Thank you, Gary.

SPEAKER_00: 1:14:09

This podcast is for educational and informational purposes only. The views express do not constitute medical or financial advice. The technologies and procedures this cost may not be commercially available or suitable for every case, and always consult with a licensed professional.