AI Model Builder for Regulatory featuring Paul D'Souza, Stuart Lacey, Harold Neto, Athena Visel

Hey!

Hey? How you doing.

Hey? Good! How are you?

You look good in black.

It's actually dark blue, but.

Oh, yeah, my wife says I'm colorblind. I'm sorry.

Find.

Hello!

Hello!

How are you?

Good to see you again.

Hi Sean!

Hey! How are you doing.

I'm good nice to meet you.

Nice to meet you as well. Hang on. I'm just being remedial.

This is the second week we have a a new. It's called Friday infocus.

Hello!

Carol is calling in from Sao Paulo, and.

Harold | LABRYNTH: Hello! Everyone nice to see you.

Is in Burundi, or someplace like that Burruni.

No, he's in the Bahamas, but I never got his.

Bermuda, Bermuda, Bermuda. My friends.

The other B.

The other B.

Yeah. Bangladesh, yeah. The place, the place of the pink sand.

Good to meet you, Sean, glad that you could host us excited to work with you today.

Great. I'm glad you guys are here. It's nice to meet all of you hope to talk a little bit more later. I'm just to start this thing and promote everybody to panelists. So just give me a second to start promoting the attendees so that they can participate as well, looking forward to your presentation.

Should I just quickly test my screen share so that it's working.

Sure.

Great. Can you see that.

Yep, Ronnie, I'm gonna stop sharing.

Bye.

Cool.

Good morning!

Good morning!

Gosh! Bruce is here!

Hey? Bruce is always here when he's not in Oklahoma.

I know right, Arlen. Good to see you.

So why don't we.

Thank you.

And let anybody that's new to the call, introduce themselves. While we're waiting for me to promote everybody to panelists. Someone can just start talking. For example. I don't know anyone.

Well, I'm I'm not never been shy to talk so. Hi, everyone hopefully you can hear me. It's Stuart Lacey, and I'm not aboard a A space shuttle like a faizan is here, so I've got a little bit jealous about that. But I am calling from Bermuda, so I can't complain.

Hi. My name is Athena Weisel, and I'm joining with Paul and Stuart today.

Awesome morning. It's good to see you, brother. We haven't spoken in a long time.

And Mark, do you know, Bruce, you guys need to talk a lot to lot to talk about? Bruce installs the clean rooms that you bring into compliance.

Oh, is that right?

Yep.

Oh, that would be interesting! Yes. Where are you located, Mark?

I wish it was on a 3,000 acre ranch like boy riding horses every day. Saddle up dude ranch kind of thing. That'd be nice.

Mark tell you, 2, 3,000 acres is that.

No, I just. I never saw.

He, he actually says 6,000. So I had figured he was lying.

Again.

Okay. I didn't say that that wasn't me. Okay, that's.

I just don't get any respect anymore.

Bruce. It's it's nice to meet you. It's it's interesting because I I've I've known for a long time since I've been in medical device and pharma and stuff I've been working. You see the gray here. I've been working quite a long time in the industry. I have noticed in putting it together, quality management systems for all different types of companies, small, medium large, that the quality management systems, that they have the trackwise agile, the master control. All those places do everything but track fully. Environmental controls. They have limb systems that you can put things like they call em testing, which is just, you know, checking the equipment to make sure that nothing was growing, and

you know they can prove the products good. But they don't qualify the space that the equipment is working. In. In other words, the clean rooms that you're building or putting into place, you turn it over, and the companies have to maintain that area. Well, nobody's got that digitized yet. So I'm I've got a crew that's putting that together for the 1st time. So we're very close to having that. And we're very interested in going to market with companies like yourself that are putting the clean rooms into place because we have another company. We've talked to that, said their customers actually told them. We want you to digitize all these paper forms, so

we're very close to getting that done, and we would certainly love to have a chat with you. Show you the program that we're building almost done very close to getting it done and completed and validated so that we can turn it over to companies like yourself. Offer this to your customers, so that instead of having paper forms. they can have all their information right then and there, and they can pull it up anytime. They want.

It's a win-win-win.

Yeah, that does. It sounds like a great idea.

And and that, my friends, is the the reason we all come back on a Friday, because you never know who you're gonna meet.

Well, it's funny. I saw the AI. I saw the AI thing, and I said, You know, somebody told me that I need to add AI into the program. And I wasn't sure how we'd be able to do that. So I said, Yeah, let's. oh, I thought, it's because you.

You saw my name there, but anyway.

Oh, I did see that I was gonna talk to you.

Sean, go ahead.

You sold me a vacation.

Sean.

So before before we get started with today's presentation. As many of you know. This is the the second Friday and focus call, which is part now part of Medtech leading voice Exchange or Mlvx, which is the new group that I've started for product and service providers in the med tech space. Where your expertise matters. So what we're doing is we're recruiting the 100 founding members. I'll put information in the chat about how to join, and towards the end of the call we're going to introduce a new segment called hopefully. We're going to introduce it. If I get the prompt called steal this prompt, and it's an opportunity to share

large language model prompts, chat, gpt prompts that you can use to accelerate your business. And so that's 1 of the new features I also would love feedback on. If you would like to have a featured Demo, if you would like to have a featured startup. If you would like to discuss a recent 5, 10 K. Different segments that we can add to this, besides just having a speaker every week, and you know a little bit more like Good morning, America. so to speak, but not quite as stupid, and the so feel free to share your feedback with me, contact me, set up a meeting with me if you're not sure if mlvx is for you, happy to walk you through it privately, anyway, that's enough of an infomercial for Mlvx. I apologize for that and want to turn it over to Paul to tell us what we're gonna talk about today.

Awesome. Thanks. Hey, guys? By the way, Sean, I just want to say. thank you for stepping up and congratulations on what you're doing with this group, and how you're rebranding it. Love everything you're talking about. So I just want to say just a shout out back to you. So thank you. Guys, I've been on this call for a lot. I've known most of you. What we're doing today is we're sharing an idea. I'm working with labyrinth now. And we're doing some fantastic things in this space of AI in today's call. We'll go into more detail is really to share with use this tectonical shift, this movement that is going on in AI, and how it's about to really impact regulatory

labyrinth. We've been with them now for 2 3 months is focused on. not just FDA, but regulatory in general, and we seem to be moving towards government and energy. When I was at the 10 XI met a lot of you, and I told you what we were doing. I did that that fast, whatever it's called, and I did explain it, and with all the conversations I've had that time with Athena for the last few 3 a couple of months. Remember, Daniel Beretto came on this call with us with one of the startups 2 months ago, had a lot of calls with him, and what we realized that the demand the opportunity to bring for AI to impact what we do with the FDA, so that everybody wins right. The FDA gets better or gets companies with more compliance. We, the consultants, we, the business leaders, can actually bring an acceleration with efficiencies that AI can bring during the process, because in our business

lives matter people die. Time is important, and the regulatory requirements go through the entire life cycle. We believe there's something here for us all to to do in reference to leveraging AI. So with that I want to introduce you to the team. We have Stuart here. Who's the CEO of the company called Labyrinth, which is the the what we're talking about today. Harold is the CTO. And and Athena is our subject matter expert? Who runs an Eqms company with that? Why don't we we start up with with Stuart. Why don't you say hello and a little bit of background, and take it from here?

Great. Thank you, Paul, and everyone. I appreciate you making the time. Today. I know how busy our lives are. So I will be as efficient as I can. And I wanna just thank also, Sean. And this group for hosting us today. what we're going to share with you today is a platform and a pathway which we believe you need to know about, because there's an opportunity for each and every one of you. Should you be interested to participate in it? And that's something called foundry. And we're going to be talking a little bit more, and you'll see why foundry is what foundry is, and where there's an opportunity and what it means. But let me start a little bit by just giving you some background. So I'm going to share a screen with you and take you through it, and I'm going to be talking through part of it. And then Athena Vizel, who you met briefly as well, will also carry part of it. But we are in the process of launching a new platform called Medtechteam.

and it is the future intersection of regulatory AI and medtech and kind of for today's agenda. Aside from the introductions we want to share with you briefly, a video explainer and what we call the ontology overview and ontology is a design workflow model of how regulatory stuff gets done efficiently with clarity and with data driven decisions. We're going to show you that we're going to give you a brief background of what brings us today. And then we're going to share with you. Why foundries being launched what it means and the role that can be played with medtechteam. So just by way of background, I'm the founder and CEO of labyrinthai on the right hand side. We are a spin out of one of the world's largest companies called invisible. If you don't know invisible you should. They are the large language model trainer behind 80% of the world's biggest models. So if you would think you know, Sam Altman and Chatgpt basically wouldn't have existed without invisible training, their model in the back end. They are the behemoth.

and they're invisible, which is kind of what their brand is is kind of. They work in the background to solve the world's largest issues, the most complex issues, especially where there are requirements for modern AI they have launched their own entity called Infinity Constellation, which is a launch platform founded by invisible to incubate game-changing AI services platform companies that are solving real world problems. And one of those launch companies is labyrinth, which I lead. and we're backed by infinity and and invisible, and we are the bridge between AI driven regulatory compliance, intelligence, and human in the loop.

And we're doing a tremendous amount of work already with government and with leading companies in infrastructure data center AI and energy. One of the reasons we're here today is that we launched at abundance 360 only about 60 days ago. and during that launch, as you can imagine, those of you that might have been there or know Peter Diamandis. They've got a huge and very powerful following in longevity, healthspan lifespan. and we were launched on the main stage, and I was swamped immediately. And then, in following discussions with Paul and members of medical Med. Tech, you know, Pharma and medical device, we suddenly had a very rapidly growing pipeline of medical interest in Med. Tech, although we hadn't yet launched that as a service offering.

And so we looked at that with a lot of interest, and at that event I met Athena, who's on the call today, and we forged a partnership immediately, and we've been collectively discovering. What do we do with this incredible interest to effectively deliver an AI platform like labyrinth, but not necessarily for government and for energy sector, but maybe for medical and medical related sector and the regulators that they work with. So that's kind of the background to give you a feel for what labyrinth does. I'm going to quickly play for you the demo, but it should give you a pretty solid idea of kind of what we bring to the table.

Oh, I'm just gonna pause it there for a sec, because we're gonna look at this in a minute, so don't feel rushed to look at it. But what you're about to see in the remainder of the video is how we deploy for permitting. Now I know you're not a permitting group, but I want you to imagine you were to just swap out any medical regulatory use case and workflow for what you're about to see, and that'll give you a feel for it. This one. okay. cool.

Did it? Did I just see that it was based on Google? Or was that my Mac? Did? I.

Natalie. I'll answer that for you right now. Great question. Thank you. So we can plug into any number of Of the I'm just gonna stop share. So I can see you. There we go. That was Nelly correct to ask that question.

Yep.

We're gonna do a full Q&A at the end. But I'm happy to just jump in on that. We can plug into any one of the models, and so we do do a fair amount with Gpt. But that's our own custom interface, which is a layer on top of models, and we are a multimodal multimodel company. And so depending on the complexity. And an example of that would be where one of our clients is a nuclear company right now. And they're going through Nrc, which is a very robust, as you can imagine. Set of requirements. Those documents have a context window, which is so large because it's so much text that Gpt is not the right model. So we would actually port that into something like Gemini, which has a much larger context model, and then chunk and then use custom built AI agents. So we are multimodal and not strapped to any one model. It based on the use case. Okay.

so just to quickly go go back. And so the reason this is so powerful is we've taken that we've built what you saw there, which is called this ontology, which is that market map. And I want you to think of this as every step in a regulatory process from beginning to end. So this could be a 5, 10 K dhf. Iso, 14, 9, 71. Whatever that requirement is, and effectively the model that orchestrates in the middle maps that entire workflow a little bit like you might imagine a subway map of New York or London, England, and it's got all the stations, all the stops, all the tracks, all the things that happen in parallel and separately, that bring you from 1 point to your final destination, which is an approval.

and each and every stop on that that kind of subway map or on that workflow ontology we can map as to time, treasure, money, resources required dependencies user judgment. Delegations, escalation sign offs approvals and they can be run through a risk based filter. And the result of that is number one, a market map that's data driven of an entire workflow. And then number 2. The ability to use our dials to optimize. optimize for time, optimize for cost, optimize for resource. And we're doing this live right now with our clients in energy accelerating significantly. Time to output and to give an example, our current nuclear client

is comparing themselves to another nuclear reactor client called Nuscale, who's 500 million dollars into their filing with the Nrc. With 350 engineers. and we believe we're going to do it at our current client. Deep fission for 50 million dollars and about 35 engineers. So that is a 10 X improvement in time, resourcing cost and acceleration of business. They've also got a 2 billion dollar revenue client already signed and announced publicly. So we will bring them to 2 billion dollars of revenue much faster than otherwise, which, as you can imagine, is powerful for patent IP competitive risk, market access, etc.

So what we want to talk to you today about is, what would that look like if we created the labyrinth of Med tech? And that's kind of being working title medtechteam. And we want to chat to you about why we think that's interesting to all of you. And what foundry is. So where we want to talk about today is that today is much more than the demo we just showed you. It's an invitation, an invitation for you who are incredibly ex experienced in what you do, who understand the own ways. You've had to manage these challenges. who've seen the system work and fail, but can now be in position to potentially lean in into a rising tide kind of lifts all boats approach.

In a co-creation fashion.

And I'm going to pass it over to Athena now to talk through the next few slides, and I'll just move the slides for you. Athena. Go ahead.

Yeah. So the problem. I'm sure we could spend hours here talking about the problem. We all have our own experiences with the problem. But really, right now the system isn't broken. It's becoming obsolete. And it's because we have so many different tools and systems. And it's becoming very fragmented. And really the the root cause of a lot of that is because they're human run. There's a lot of interpretation that has to happen. We all have different experiences. Everyone in the company has different experiences. We've all had different reviewers. different inspectors, different leadership and different expectations. So everyone's prioritizing things differently. Everyone is bringing in their own war stories, and where they've been burned, and sometimes that starts to creep away from what the actual regulatory expectations are and what the fastest line is and where we need to go.

There's also just a lot of manual effort. So even with all the tools they still require it, they're only as good as what you put into them, and then they're only as good as whoever's monitoring them. So there's still a lot of manual input that's required. And there's also it. It really relies heavily on the people that you have involved. So if someone leaves, you could lose like an entire set of knowledge with that person leaving. you can go ahead and go to the next one. And so that makes this. This is not scalable, especially when you're relying really, really heavily on human capital. It's not scalable. So the example that Stuart was just giving

with large companies and the army that they need to get things across. You're probably all familiar with the Venn diagram. You can have it cheap. You can have it quick, or you can have it with good quality, but rarely can we get all 3. And oftentimes we can't even get 2. So a lot of times we're running in very lean organizations. We have objectives. We're doing everything as quickly as we can, and a lot of stuff falls by the wayside, and we have to live with the mentality of we'll fix it later. and then later comes usually when you're preparing for submission or you're preparing for an inspection, and then it's a fire drill, and then you're trying to piece it

the pieces together. You're trying to do all the cleanup you've you've had turnover since that point, and it would just wouldn't it be amazing if this stuff was managed from the beginning with an intelligence built into the system, and because of this heavy reliance on people, it slows down the process. We get a lot of lost time with rework, a lot of lost time, possibly with rejection of submissions or with inspection findings, and it adds to a lot of cost. So go ahead and go to the next one. And so the shift that we're looking for is really we don't need more tools. We have a lot of tools, and a lot of our tools are highly over engineered, like they have so many workflows. I have clients that have to take out the workflows because they just don't have the manpower in order to operate those workflows.

And so we don't need more tools. We don't need more templates. We need intelligence built into these systems. We need support for the experts that are working in these systems, and also to take over some of the I guess you would say, lower level, like Qc. Tests, or just keeping us keeping the guardrails on the experts so that we don't have to hire like 7 document control people. We don't need to be relying on our engineer to also be a document control manager. That's where the building in the Agentic AI and having the ontology built into the system to support the systems that are already there and to support the humans, we're not looking to

completely take humans out of the equation. We need the humans, we need the humans there to do what humans are still the superior being at doing. But we also want to support them, because I'm sure you're all aware we have a lot of burnout, a lot of turnover, because there's just so much expectation that is relying on running these regulated processes. And, Stuart, if you want to go.

If I could, if I could just add, for those people who are dialed in, if you're not 100 sure about what a gentic AI is. Let me just quickly explain that because there's so many tools and terminology out there these days, we've explained ontology, which is just a complex workflow that's automated and mapped. Agentic AI is what has really become the biggest story in late 24, and all of 2025, and what it is is the maturity of all of what you've seen with Chat, Gpt and Grok and Gemini on top of what was called deep research. Which are these reasoning models that could use what was called chain of thought to duplicate what researchers were doing. And so suddenly, we had this incredible ability to go more than prompt to actually do incredible research.

Agentic AI is the next layer on top of that. That gives autonomy to the degree you decide to an AI agent to actually follow a sequence of steps using these AI tools so that might look like I have an agent that will book my flight for me after I decide where I want to go on vacation, and it would know how to reach out to expedia or kayak and pick the flight in your seat and pay for it, and everything that would be an agent or an agentic experience. And so what agents look like in our world is? There are sometimes agents that can automate a process just like filing. It's done. It's signed off. It's approved. You need to file monitor for approval and monitor for any feedback, or questions that might be fully agentic.

And there are a lot of steps that we in labyrinth world called Minotaur, because in a labyrinth you have to have a Minotaur, and that's half human, half beast. And that is where it's half AI, half human. And what that means is what we talk about human in the loop. That is where an agent cannot act fully, autonomously. But it'll do 80% of the lifting with the humans doing the 20% of the oversight, judgment, decision. So when you see ontology and agentic, that is a combination of workflow and agents working with human in the loop to solve the problem. And that's how we get the clarity and the velocity.

So maybe, Athena, I'll continue with the operating model. So I think actually, I just did, I probably should have gone to that page. There you go.

That's.

Of that. Okay, there you go. Ontology, structured logic rules, roles, agentic, executing, flagging, and escalating, and the human scalable judgment. Now, I can tell you it's probably not 33, 33, 33, I would say the ontology, designing, building and mapping. It is the most difficult upfront, because we have to train models and train the system. The agentic AI is an output which is very quick, but has to be coded, and the human in the loop, I think, is running at around 10 to 20% of the time. But what you'll find most interesting about this is, if any of you have heard about AI models, or how you train a model

once you've trained a model, you do not need to train it again. So if you look at like language translation. Once you've trained a model in Spanish, English, it knows Spanish, and if if it misses a local dialect the one time you train it in that additional local dialect. It is now trained on that it would never be trained again. So the models improve. the scalability increases, and the marginal cost of use drops every time they're used. So think about a living, breathing, learning system that every time this group would interact, or customers or clients or regulators, it's learned from the last instance.

It's smarter, enriched quicker and has a higher roi, because the input cost to get there is lower because other people might have trained it to the point you use it.

But how do you calibrate? How do you calibrate it? So you're saying it's learning from other users. But as we all know as a team here as professionals. There are people on your team that are users that you don't want anything to learn from.

Correct. And so your instance that you're running in Nelly. What I'm gonna ask for people is not to single you up. But if you want to put your hand up quickly, then I'll try and go through all the questions, because I want to get to the Q&A at the end. But you make a great question, Nelly. You effectively as a user, use a private instance of the model, the then latest, best model, just like you use word, the latest instance of Microsoft word or excel, or Powerpoint. You're using it with your data connected to your vault with you. Decide to use it, not us. We update the model. And then, when the next version of the models released, your instance gets to use that next version so similar to any software you might have used. But

but kind of does that help answer your question, Ellie?

Yeah.

Okay, great, perfect. All right. So Athena want to chat about some of the use cases.

Yeah, there's a number of use cases. But we just wanted to include a couple to get the imagination going. So 5, 10 K. Readiness having workflows and assistance set up for getting quick to quick to the end goal. So also putting in initially like, this is what we want to do. And it's setting out the pathway for us, also helping with the submissions, making sure that we have everything I mean, even with the Easter. If you don't click one of the buttons. You might not get some of the questions, so the support that would be available to help get your 510 K submission ready. Also managing divine design history files, linking inputs to if you make changes to downstream

inputs and how that will impact upstream. So that oftentimes where I can see a lot of gaps happening. And so something that's managing that process risk management. That's quite the spider web and a change over here, if not paid very close attention to impacts 3 things over here and having an intelligence built into the system to support the people working in it. To say, you need to check this. This has happened. This is going to be impacted. Audit readiness. This is something that really aligns with where I focus. And that is also trying to stay audit. We all want to be audit ready at all times, but we also know the reality, and we know the deficiencies with the systems we have now, and helping keep us that way, and so staying audit ready at all times, running micro audits and then identifying in real time where there's higher risk. So also helping us with that risk, determination on what needs to be addressed and then post market signals is another example that also ties into risk management. So

dealing with like the complaints, the anything that comes in after and how that's going to impact and impact updates to the design going forward. So those are just a few example use cases. But really the list is can be tailored to your organization and is is endless.

I I see a question from outer space from Faizan. So while he's got connectivity from a.

So good. I get to talk to the International Space Station.

Thank you. Everyone. This is all pretty exciting, and it's very interesting. I wanted to ask, like, what type of file formats can the AI analyze, because so far, what I've tried with like chat, gpt and deepseek. They're like some file formats like Pdf, or like word, they work fine, but when it comes to like excel, or something it might like, you know, just divert to somewhere else. And I wanted to ask like, is it also like trained on design files like 3D models, basically design files. Just a thought.

Yeah, Harold, do you want to jump in and maybe share some concepts on that.

Yeah, that's 1 of the reasons we're using the agentic workflow because we we use different models for different tasks. So some models. We can use to analyze the tables, the excel files. There's also new techniques that where you transform those tables into a more like Llm digestible concept in terms of design files. We're still testing a few tools for 3D. Right now. It's not the best use case. That's 1 of the moments where you need the human in the loop so far. And it's also for the moment very expensive in terms of AI use to analyze images. Okay? So on the data and number side. That's okay. That's covered on the image and design files. Not yet.

And to give you an example, that one of the things that's really powerful about the tool is if you are going through an application process where someone has gone down the road before you like. I always think of Ponce de Leon with the the machete going through the underbrush of Florida much easier to walk down the path after he's done that, or she's done that cutting first, st right? And so the great thing about the way that the systems work and the way that we at labyrinth but would be the same approach for medtech team is, we can train the system on what great looks like from precedent and citation, and then it can pre-complete to a very high level.

a draft application, draft, submission draft whatever using precedent, using language, using citation and then flagging with data tags where it doesn't know what? Specifically from your proprietary data it requires to complete. And now all your people are doing is out of 100% of their time. They're using only about 10 or 20% to do that review and sign of completion rather than doing all the table, build, document, build document creation. and they're using our tools which are trained for this rather than your own. Gpt, which you're correct. If you throw a spreadsheet or something to that. There's a lot of hallucination, a lot of risk kind of doing this as a home kind of, you know, practitioner. And so this is now where professional AI is being used because we can train custom agents for every single type of workflow. So great question. Thank you. I'll pause there any other questions before we continue. We've got about another 5, 8 min to explain where we're going and why, and then we're going to open it up to more Q. And a. But I just want to pause. If there are any other questions

good, I'll continue then perfect. So if we just pop back up to the document. You'll see we're talking about the use cases, and we want to introduce foundry. So foundry is an approach that I've used many times in spinning up businesses that I've had in the past. and it's a concept of pre-populating an elite advisory board prior to the launch of a new entity whereby that elite Advisory Board acts as collaborators who help shape and ideate what the platform might look like. Where they help define. Maybe what the ontology and regulatory language and expertise might be useful for where we kind of move from what would be a siloed approach to a shared more distributed approach. Where there's a collaborative and collective intelligence. That's what a foundry is. It's a pre business launch entity or advisory group.

And what foundry members do is we'll probably have bi-weekly sessions or twice a month to learn. Listen and help. You see what's happening in the state of the art in the business. And so one great thing is, if you just might be someone that likes to see what's happening and get informed and learn. That's kind of where your value statement shows up. I want to see what's happening, understand the state of the art, and then take away and apply it. That's fine. We would love, though obviously, for part of that join is for you to advise on what you're seeing. The problem, state the opportunity we are, and you might be sharing insights and real world processes and learnings as we go. There would be collective voting on roadmap features and priorities, for the system

would be validating and refining effectively the 1st version before it ships, and you would have utility and visibility into the pre-launch version. which is a big value. If you think about that in terms of maybe some of your own businesses and products and services, and you'd shape not just the baseline foundation, but also help shape the features, and where you find value as a user experience. So think of this not as a pilot, but participation. But, more importantly, we want to make sure that there's upside. And so we will be taking a slug of the launch equity of medtech talk team and allocating it to every single member and foundry.

Okay. Now, to give you an idea, this company will be backed with a minimum of 5 million dollars cash on launch. And I'll guarantee to that. Okay, so the goal here is not to raise capital from you. Quite the opposite. Okay, it's to ensure that we have the financial means to hire the correct engineering team go to market team and leadership team to deliberately execute on the plan, and foundry is an advisory board to it. and we will allocate a portion of the equity of medtech team to the members of the foundry team. Okay? Number 2, early access to models. You will be on the inside track, then, of what we're building, of course, subject to an nda. So that's 1 of the requirements of being on foundry, and then through that have visibility on what is happening so that you could potentially get on the inside track of launch and accelerate some of your businesses.

Strategic insights into AI infrastructure. What you don't know in this world can hurt you. What are the risk? Mechanisms, parameters, copywriting, IP, labeling, sorting, scraping what is permissible? Not how do we do it? How do we do it? Well, and we'll be learning those from the cutting edge as well of all the work we're doing in energy and government. and then upskilling part of the great thing about this is, some of you might be black belts already in AI. Some of you might be white and yellow belts, and this is an opportunity for you to personally, Upskill. I see a couple of questions. I'll happily pause now and take them. Arlen. Your hand was up first.st Go ahead.

Yeah, thanks. So in this foundry team, what if any role is there for physicians or a clinical end users. be they advisors or physician, entrepreneur AI startup, entrepreneurs.

There are effectively positions for all, but those might be defined by subcommittees. So I love that comment. Now, to give an example in the work we're doing in energy already at labyrinth. We have already Ex. Nrc. Specialists who used to work for the Nrc advising and counseling what we're doing. We have people that have worked for the Dod counseling, advising. So we surround ourselves with an advisory board, in fact, invisible. who trains global models has an advisory network of Phds and global leaders of over 6,500 people already in model training, right? But there might be some in Med Tech. But my guess is what we're going to want to do with regards to this is populate our own advisory group. And so we would probably have a lead committee, and maybe some subcommittees which would launch over time where we could bring in additional advice as required. Does that answer your question, Arlen?

So what I'm hearing is, it depends. As long as the person like me or someone like me wants to participate.

And is qualified and shows up and adds value. Yes, I mean, it's we're not going to be excited about Freeloaders. Right? I mean, the reality is, we're all in this for rising tide, and I'd like everyone to. It's like a chain. I'd like everyone to link arms and and help this thing forward. I mean, that's our invitation. Right? You know, we appreciate that we're surrounded on the school by extraordinary knowledge. and this system will be better because of that Leonard question. I can't hear you, Leonard. Do you mind unmuting? Thank you.

And I'm used to zoom calls. So that was my bad So I'm involved in standards development. And I've been just starting into chat Gpt very recently, and I just did a test like yesterday. On testing it on a standard I use all the time, and it came up with some very good hallucinations coming up with clauses that just don't exist in the standard. So the standards I deal with are copyrighted. So I wanted to know how you deal with copyrighted material like Athena brought up. Athena brought up 14, 9, 71, which is a copyrighted document. So that was one question.

2 good questions. And I appreciate you being so forthright as being a newer user. What I can tell you is, I've run a global institute on AI training and prompt training for the last 4 years. So usually the quality of the prompt equals the quality of the output and the quality of this data structure allows to reduce hallucination, and one thing we would happily do is with the Advisory Board is, take you through a prompt masterclass, or give you access to some training which would upskill you dramatically because, the a really good prompt is about 450 words stacked sequence and done under what's called a rise in framework, and we can share with you how to do that in the moment.

Okay. do it. They literally come back and go. You've got to be kidding me like. I had no idea what I was doing on this model until someone trained me how to do it correctly, and the quality of the output goes from 0 to north of 90%. And to give an example, our head advisor, who spent 20 years at Nrc. The nuclear regulatory committee. I sat with her 3 weeks ago. She had almost the identical comment we took her through training and 4 days in. She's like this blows my mind absolutely blows my mind. The difference between pre and post. So would be happy for you to kind of experience that change number one, number 2.

Copyrighted material is something that we we can talk through. there's a whole bunch of work in AI and Gen. AI around IP and everything. And so I wouldn't recommend that people in a public forum share non public IP, right? Because obviously there could be pre filing, or whatever and disclosure. or you could disclose against yourself sometimes, but with regards to the rest of the model, we have very strong guardrails on how to do it when it's acceptable and how it's done. Well, so we can share that with you again as a process. But we solve that elegantly and easily with the work we do at labyrinth. So we have a roadmap for that. Okay.

That's great. And one other question within standards. Since I've been working on them for about 20 years. Now, there are a lot of clauses that are written really poorly. And so there's a lot of interpretation that's involved in them, because they're written so badly. To put it nicely that you can get multiple interpretations. So that's gonna be a challenge within a model. I believe.

It definitely is, and I saw Nelly nodding her head. So I have a feeling, she agreed with that comment, and maybe some others. And one thing I can tell you, and having spent 15 years building Regtech, which I've done prior to this, and I don't view the Regulators as bad people. They're not a 4 letter word. They're human beings who are struggling as well.

Very much.

Integrated systems, and our job is to make them look like heroes. And so one thing I can share with you in early discussions already. And I've done this for 15 years with regulators when we bring incredible technology. In the in the early days it was like Cloud and Sas. And now it's AI, but we become enablers for them to get cleaner, better constructed, better filed, better cited. more easy to understand, and also written against what they publish as best practice, because we can ask them for that. As a result, some of the lack of clarity in the question asking and filing is actually more of an issue that they haven't updated their Reg. Not that we're not complying to their standard, and the model can learn what great looks like

which bypasses that recycle loop quite a bit. What we're hoping is that regulators start to become more AI forward. We're seeing some of it in many regulatory areas where they might actually go. Oh, my God, you're providing stuff better than our spec. Maybe we need to upgrade our spec or make our spec align with the AI. That will happen with time right now. I can't recommend to you, or even endorse, that. I have the strength or ability to influence regulatory change, but we can demonstrate by practice extraordinary clarity, documentation, citation. In such a way that we make their job easy, and ultimately we think the water will rise for the Regulators. Well, in the harbor. Okay, it's a great question.

Although I think Stuart, I hate to interject here, I think really what Leonard, and correct me. If I'm wrong, Leonard is, there are times when the Standard says something has a statement, and the industry has interpreted it one way, but it took a collective mindset to get to that standard of understanding which then leads into business best practices. So I mean, I think there's a chain of thought, and it's and I think that's where Leonard was going with. Is that am I correct, Leonard.

So

Oh, wait!

A great example is with the FDA with essential performance. And Emc. My partner and I had to actually do a training with the FDA to change their mind on how they interpreted that issue, because they were giving our clients bad information and rejecting their submissions, and after that training. Our clients were actually getting approvals.

So, so.

And we had 70 people from the FDA in that training, and.

Oh, sorry. Yeah. I'm happy to stop sharing so easy. I'd I'd like to finish a couple of slides and we'll finish up. I will answer that question. It's a good one. Let let me generically answer that which is not a full service to you. But we'll give you a fair response. Okay. I'd love to continue this conversation in one of our foundry meetings number one. So I'm not trying to write it off. I'm saying. Okay, number 2 I would submit the following, that Leonard, you and your partner, and the people working this, and Nelly, who mentioned everyone else nodding. are trainable like as human beings. We learn right. We learn from experience, collaboration from partnership, from precedent, and we adapt, and everything. And the great thing about that is, so do these models learn and train. That's the entire concept of what we're doing is replicating that. And so if it's capable to be done in that respect, we can do it in the model. Now, if you, said Stuart.

we've got a bit of a challenge, because I stand by a patient's bed, and I visibly evaluate them. And I'm using empathy. And I'm using my training for micro body language and fascial recognition. And that's an empathetic thing that right now the models aren't great at. And so if I put computer vision in that room, you would beat computer vision. Okay? Fortunately, we're not doing that necessarily for this regulatory filing process. Right? We're using primarily learned systems that can update the same way your training would update. So does that make sense in a generalistic approach. Okay, cool. All right, I'm gonna quickly finish up and

So why we lean in part one, part 2, you can contribute to a legacy even that beyond your organization, shape, industry-wide best practices building an infrastructure for all a shared solution and share knowledge, not just frustration. I'm a big believer, everyone that we are all on some Laszlo's hierarchy curve to be better versions of ourselves. It's how I govern my life. I'm governed by impact. I'm governed by paying it forward. Not everyone is. But sometimes this is an opportunity. If you aren't, to actually experience that. And if you are, this is actually one of those mechanisms where we can really contribute to something that's amazing. Because in some areas, if we can accelerate

incredible life saving method methods, systems and products to market. Then that is great for all kinds of reasons, not just for the profitability of of the people that are doing it. And secondly, we're talking to some medical Vcs already who are worried that they evaluate early stage companies and don't actually know if it's realistic or not for them to get to market. And if it's actually going to, you know, what is the runway, what is the burn? What is the likelihood of actually developing this? And by creating the ontology method and model, we can actually create data driven decisions. We're actually helping cash

management and in fundraising. So we're already seeing companies go. We need an ontology before we can go raise, which is really interesting. Because now you can make data driven decisions about what it looks like, which also helps on the investment and balance sheet management side. Real momentum. We have about 15 plus companies in the pipeline that want to do this, which is great. We got customers waiting for a product. They will be ported from labyrinth over to medtech team. It'll be an inbound customer, pipeline for the growth of the company. At recent events where Paul has attended and we've spoken, we've had more than 4 out of 5 people on average express interest in learning more being involved, supporting it or becoming a pilot customer.

And Regtech and AI is in the top trending searches online, which means there's massive interest in the space. And there's yet to be strong solutions. Athena Paul, do you want to comment on Eastar and everything else? Quickly.

Yeah, Tina, why don't you be there? Yeah, it's just the timing of.

The timing right now is, of course, impeccable, because we have a lot of stuff that is happening Easter as an example, but just the modernization, where also the FDA is really embracing AI. Everyone is learning together, and we want to move forward along that path. So to what Stuart was talking about earlier, even how what we're doing can be influencing, how the regular regulators are comfortable, moving forward with certain aspects of processes, and where they will be focusing as well.

One comment I'd like to make on top of that is that for all of us here on the call that if you're a consultant, and if you're trying to do more in the regulatory space, this could be a tool. Consider what we're building out a tool in your toolkit to help you do your work better and faster, like you know, Mr. 6. 0 6. 0 1 guy! Right, Leo, like you could really lean in and help us all for that regulation. Be the expert, and then leverage this platform. When you walk into your clients as well, we'd like to explore that, and that's the invitation. That's what we like to get out of it today.

So it wouldn't be very helpful if I'm telling you all this, and we couldn't prove it. So. Any of you. This is just within the last week. the the platform. Infinity constellation is being called the Berkshire Hathaway of AI services. They've just raised another round 17 million. This is literally the cutting edge of what is called Sas version 2.0 instead of software. As a service, it's service as software. And AI service platforms are literally the future of how technology will be delivered, and so don't trust us, but feel free. We can share the links afterwards. But there's a whole new deployment mechanism.

and if I shared with you at the beginning, I'll just go to the top. But the actual companies behind it. So the entire platform is back by invisible. It's launched on the Infinity Constellation platform and the new entity, just like labyrinth medtech team, if we call it that, it's a holding name right now will not only have launch funding of 5 million dollars to get it up and running, but we'll also have the power of this group and finance marketing. Hr. Legal and everything is an integrated service offered by infinity constellation, just like I do at Labyrinth. I don't have any of those team members. I don't need them. It's leveraged through the group. What that means is, medtech team will be launched just with product engineering and go to market, which is a very powerful way to launch new businesses. So if any of you want to look at any of these articles as to why and how

this is the way companies are getting to revenue. And to give an example, the most recent infinity constellation business went from 0 to 10 million arr in 6 months. That's 10 million of annualized revenue in 6 months. That's a hundred 1 million dollar valuation in 6 months. That's what's happening right now. So next steps, and I'll put into chat. But there's a type form which we would if you're interested in joining and just learning more and becoming part of it. I'll put this into chat for you. I can't share with everyone. I think I can only share with hosts and panelists. How do I share with everyone.

Maybe one of the hosts can share that link again wider. But it is this link.

Can you share it?

Thank you. Sean.

If you send it to me, I will share it.

You should have it in the chat, because if you're a host, you should see it there in the chat window, and in fact, forward it to everyone. So what we've created is a type form and the type form literally just says name and contact information. So we can reach out to you. We are not asking for money because we have it. We're not asking for too much time, but we're asking for some, and we offering an incredible opportunity to be part of the foundry. and that window will close. June 15.th Medtech expects to launch July one. and the 1st meeting of the foundry will be the second week of July, probably the 16.th That's our timeline.

and with that any Q. And a. So please. Thank you, Sean, for sharing that. That's the type form link you're not committing on the type form you're just flagging. You're interested, and then we can reach out to you and go through a little bit more detail everything what it looks like ideal numbers. We're looking for about 15 people to be on foundry would be ideal from a diverse background. Okay, I will stop, share and pause, and just thank you all for having taken the time to listen and answer any questions you have with the time left Shawn. I see your hand up.

Hi, sorry that doesn't allow me to raise my hand electronically.

Okay.

Most so my question for you is, is, why med tech? Why, now.

Why, Medtech? Because last night I was trying to come up with a domain that I could buy on Godaddy and Medtechteam showed up, and I grabbed it for my $49, and so we can call it whatever we want. It's a placeholder. But why, Medtechteam is medtech is like Regtech. It's the field. We know that Andteam is because it's being put together as a team of agents and a foundry team to launch. And so there is intent behind the naming, but that's what it embodies, but we can call it anything. For example, labyrinthai. We could call it anything.ai. If you want as well. What was your second question, Sean?

What now? My my question is, why did you choose the medtech industry?

We? Oh, it's simply because if I went back to that page above when we went out to market and we launched an abundance, and then, you know, a pool went to some of the sessions, including here. We've been overwhelmed at labyrinth with inbound medical customers that we aren't currently taking because we're focused primarily on permitting licensing energy and state Gov. and as a result, we have an entire flow of clients and business and opportunity that we're not actually in a position to immediately support. And so we're launching a complete sister company to manage and support all of that literally off the back of what we're doing. So it is. It is to bring this value to this space

quicker through a separately funded, entirely, separately running entity. That's kind of why.

And and also just to add, on top of that, because I think the focus is, it's so deep and wide that it's very different to nuclear, if you would right. So regulatory just warrants its own attention.

Not just regulatory. You've got regulatory and quality. Everybody thinks regulatory submissions, only you've got an entire vast pool of just quality management stuff that you.

Correct. And that's my point.

And then you look at the look at what? And you know. And I really made the decision and my feedback. After I went to 10 x guys was that we should bring this and fund it and do it separately. That was my recommendation, after all the conversations I had with you guys. because what we realized it was so deep and so wide, the opportunity was so big like, I spoke to Daniel Beretto, who was on one of these calls right? The Ex FDA guy. And he was just rattling off all the things. Tina is an industry expert. She focuses on quality right? And then she realized how much it was. And then, Leo, you educated me on a ride to the airport. Right? It's like, Wow, that is like, I just realized. Like I it is. I did not know what I did not know.

Yeah, worthy of its own, its own brand and.

And to answer Nelly's question, Eqms and Qms. Is a big part of what we're expecting. This new company will do. You want to comment on.

Yeah, when I met Stuart, I latched on to him because I've been dealing, you know, dealing with this for years, and just seeing where their intelligence really needs to be built into the system, and how often I brought in as a consultant, and the support I would love for these systems to give. And when I was learning what invisible could do and what the possibilities are, that's why I really latched on to Stuart. And I said, we really need this in our industry.

Yeah. And and, Nelly to your point. The foundry is going to be exactly discussing what is the wedge product? What is the initial launch. What is the demand? Where do we go? How do we do it? How do we? Those are all the kind of questions. So I love. I love your question. It's the right question. Okay? I see Leonard, Jose and and Faizan. I'm going to go with Jose first, st because he hasn't spoken yet, and then I'll try and get the other 2 right after. Go ahead, Jose.

Sure, thanks. I'm just curious. I mean, you kind of state of the 10,000 foot view of conceptually what you're looking at, and I think a lot of people think AI, and then just think, chat, gpt, and a chat interface. But I wonder if you can speak a little bit to kind of the ui that you're envisioning for this. Maybe what you've done in the other industries and like, what is that interface and and that interaction between the user and the tool that you're envisioning.

It's it's a great question, and I wish we had time to do a whole hour on that. And Harold could show you what we're building. But to answer it simply, if any of you have been on a a platform called Manus MANU. S. Feel free to go have a look at it. The concept of manus is like a more modern Ui interface. Where, on the left hand side, it feels more like a Gpt, you're asking questions, prompting learning, and on the right hand, side agents are working autonomously on the work that's being done. And so you kind of get a split window of Wow! Here's my research agent kicking in. Here's what they're doing. Here's my validation, my verification, my asymmetric. So it's kind of like a that's the kind of feel I think we're aiming for. But again, we need to meet the requirements of the

the the customer for this product, and if they want a certain interface. That's the despite design spec. We'll we'll we'll build on. Harold is a master. Lead on product and system architecture. He's worked with Amazon, Cisco, and stuff like this. There's probably one of the best in the world and product design and architecture. So the Us. Will be designed for the customer requirement and the value proposition. Leonard, and then faison.

So 2 quick things.

Yep.

One is another area to look at. But I think it's a much longer lead is the clinical area, I think, is gonna be something that's huge as well. And another is that regulatory database that I work with a lot that that has done their own. AI is a met board just to look at what they've done potentially, and I've been working with them for a long time.

So love that maybe an introduction second one. I love 1st one. We're ready on it. In fact, Peter Diamandis's entire longevity team is focused on clinical, and we're in discussions with them already to maybe maybe be part of this. So thank you for that team. I've got another call starting right now at one o'clock, but I've got 1 min, and then I'll let Sean wrap up. But Faizan, quickly to you.

Okay, thank you, Stuart. So my question was, what's the expected output of the AI because to give an example of when I was developing procedures for our Eqms. So what I will do was, I will give it clause of Iso 1, 3, 4, 8, 5 and the relevant regulations, and then I will ask it to generate the mermaid code so it can like create a flow chart by using the code. and then I will use the code and the separate mind application, and then I will have like a whole flow chart. So instead of like starting from 0. I'll be like starting from like 80.

So. So what I might propose is, maybe Harold and Faison can have a 1 off chat about that. I know people are dropping, and I don't want to let anyone, Miss Sean. I'll let you close out. Thank you for letting us speak with this group today over to you.

Thank you very much. Thank you for attending. Run to your next call. Don't be late and appreciate all of the enthusiasm and participation. Today I would like to share that next week we will have See if I can find the right screen. Guest speaker will be Ignacio Smet, CEO, and co-founder of eon minds who will be presenting his platform for medtech patents. So you won't want to miss the medtech patent event. and I'm looking forward to that. I put a steal this prompt in the chat. If people can see it, I'm not sure. Let me know if you can. That's a way of introducing yourself a way of talking about yourself. Those of you who, like me, have a little difficulty talking about yourself.

This prompt is designed to help you identify your expertise, bring it to the front fore and talk about it in a way that doesn't make you sound obnoxious. And so the goal here is that you all think about joining. Where is my screen? stop stop share right? Stop sharing. So yeah, please join us next week and bring a friend. I would love to have conversations with anybody on the call about joining Mlvx. We're off to a great start, and the platform is going, live today. So enroll, enroll, be one of the founding members, and thanks again to everyone who participated

and showed up today. And we will see you next week.

Thank you. Everyone.

Thank everyone.

Thanks.

Thanks, bye, bye.

But.