Julian: Hey everyone. Julianwith Behind Company Lines here today with John Cordier, CEO of Epistemix.Epistemix enables organizations to improve forecasts and manage risk by usinginteractive synthetic populations to test strategies impacting health,economic, social, and environmental outcomes. John, I'm so excited to chat withyou.

Not only. Because of your founderjourney, but also, Epistemix and what you're doing and, and the ability reallyto have this whole testing environment that doesn't actually impact people andagain, still gives accurate results into whatever companies kind of choose totest. So really excited to get to what that means for the future of testing andtesting different types of of scenarios and situations and, and, mm-hmm.anything in, in a population. But before we get started, describe to theaudience what the heck is a synthetic population?  

John: Sure, sure. Well, gladto be on. Thanks for having me. Of course. So the synthetic population in ourcontext is a representation of every single person, school, household, andworkplace in the entire United States.

So some people might refer to this inCommon Talk as a digital twin of the entire population for a more technicalminded person. What we're doing is we're taking. available public data sets asat the lowest level of granularity that we can from the Census Bureau, laborStatistics, department of Education, American Community Survey data sets that alot of people are using.

But what we do with them is, let's sayyou have a zip code and that zip code is supposed to have 1500 people in it.Yeah. Well, what we do is we create a representation of a person as an. of all1500 people there and the distribution of age, race, sex, household income howmany people live in the households.

All of that is within our syntheticpopulation. The other thing that we've done is we've created a a patterns oflife model is what some people might call it. Yeah. We'll call it like ourback, the background sociology. The US Yeah. Or whatever other country we'reworking in. So within the synthetic population, without any coding, likeindividuals are already going to school if they're school-aged.

And within the schools there's certaintypes of contact patterns. And what we found is, yeah, in the lower grades,they end up interacting pretty much with their, like the people in theirclasses. Right. But as you grow out of elementary school and middle school,there's a little bit more interaction between grade levels and high school.

There's a ton of interaction betweengrade levels. Same thing can happen in the workplace environments, depending onwhat the workplace is, number of people. You have a different number of peopleyou might bump into during the day there. And specific to our synthetic populationsinitially we were in the the Graduate School of Public Health at the Universityof Pittsburgh.

Building this simulation platform andbuilding synthetic populations to understand the spread of infectious diseases.Yeah. What we've adapted our synthetic populations to be able to study todayis, doesn't have to be a disease, but like other things that transmit betweenpeople. Yeah. Like ideas or products sentiment, like all those types of things.

Sure. You can begin looking at from aninteractions between people interacting with other people. Or as theenvironment changes around a person, how does that influence that individual'sbehavior? Yeah. And that could be both like your physical surroundings and alsowhat information you're being surrounded with too, so, yeah.

So yeah, our synthetic population, somepeople might call it a digital twin, but like we look at it as a really richdata set that enables people to start asking questions. How populations arechanging over time.  

Julian: Yeah. And how do youincorporate the behaviors that people have within that synthetic population?

Say if I'm, ex demographic, ex genderex, if I essentially have, all the variables, how do you really account forbehavior? Or, or is there already kind of a prescribed types of behavior thathumans kind of. Exhibit without having to do say any, in any, any interactionor any kind of input into that data model.

John: Yeah. So, like a userof our software, they get to define like what the behavior and describe whatinfluences that behavior. So, in some cases it could be as simple asindividuals in the synthetic population that look like. This, they like thecolor blue. We know people that like the color blue, and we have, know, theagents that would all like the color blue in this sort of situation, theybehave in one way.

Mm-hmm. . And people that like the colorblue are not as influenced by people that like the color yellow, but they aremore influenced by people who like the color red or, whatever it might be. And,and like, I'll use that as like a simplistic example. The, the use cases thatwe're getting into are things.

what makes somebody vaccine hesitant.Mm-hmm. or what makes somebody more likely to believe something they see in thenews and act on it than somebody else. Mm-hmm. What makes somebody infl? Likehow are you able to influence the decision that somebody makes to purchase aproduct through a user journey?

Based off of not only like whatinformation you're presenting to them, but how they interact with other peoplein their social network. Yeah. Other people in their, at home, at school, atwork. So, I think the, the root of your question was how. represent behaviors.Yeah. So one of the, the novel things that we've done at Epistemix is createwhat's called the Fred Modeling language.

Mm-hmm. . So you can think of this as anew programming language, right. That is specifically built to simulatebehaviors of people with synthetic populations. Yeah. So, what a user gets todo is say like, all right, the synthetic population's already there. All I haveto think about. What are the interactions that I'm trying to describe and howare those interactions impacting behaviors?

And then when you run that simulationforward, you can see the range of outcomes that happen. So, what our softwareis really good at is when there's a non-linear change in a population behavior.Sure. So, like very recently, and like a topical use case for agent-basedsimulation would be this is like newsjacking.

The, the Silicon Valley Bank. Sure.Phenomena that happened. So like everybody on, what was it, Friday last week?Yeah. Like, didn't wake up and was like, you know what? I'm going to, I'm gonnalike take all my money outta the bank. Sure. That wasn't something that waslike, their individual choice, they're influenced by other people around themand the behaviors of others.

And the next thing you know, you getthis like non-linear, like big. crazy phenomenon that happened. Right. And nextthing you know, the 20th largest bank in the United States fails. Yeah. So likeyou begin seeing these non-linear things that can happen when you're using thistype of simulation. And the cool thing about, it's like we're looking at whenthose things happen in entire populations.

So, if people like are familiar with thebook, the Tipping Point by Malcolm Blackwell. Yeah. He talks about. Thesesocial contagions that happen. And next thing you know like there was like theone shoe that talked about like it was a bunch of hipsters in Brooklyn, butthen now, then just like, took off all over the place or, yeah.

There used to be like a whole bunch ofcrime in one city and then next thing you know, just like totally dropped offand like, these are all based on behaviors of populations. and the influencesthat we as individuals have on one another. So those are like the cool types ofthings we have to look at Epistemix and what our users yeah. Use our softwareto explore.  

Julian: Yeah. And if I'm auser using the software and, and do I see outputs of say, multiple differentpossibilities within a certain event type. And so I can see kind of. If things,you know happen, say, I, I'm a comic book guy. So in a different universe, youhave your sure, pretty much your outline of different universes where decisionscould have been made in one direction or another.

Am I able to see a range of waves that apopulation is infected by, said stimulus?

John: Yeah, that, that'sexactly it. So like opposed to some like machine learning type approaches whereyou're like just getting like more and more narrow with like, this is like the., like what you get to see in simulation and using a synthetic population andagent based simulation on top of it is the range of possible futures.

That's something that we talk about.Yeah. Or the range of uncertainty. And the cool thing that you get to unlockare insights as to what trajectory. Yeah. Like you're actually on with yourbusiness, with the health of your population, whatever it might be. So Yeah.Yeah. We're, we're all about generating, and some people use like the termsynthetic data.

simulation, I think is always synthetic.Data's just like a, a way of rebranding it. Yeah, yeah. You end up likegenerating these what if scenarios that, give you insights into what analternative future could possibly look like, and then you understand whattrajectory you're on and how those different, what ifs play out over time.

So yeah, that's uh, all things thatusers get to do and they, like a user gets to define like different stateswithin the model. Sure which what they want to visualize. So a lot of times,like we visualize stuff on a map because it's easy to see like the spread andpropagation of something. Right? And when people like see their neighborhood ortheir city in something like propagating across it, they connect with thatmuch, much easier than saying like, here's another graph that, you know, orhere's a scatter plot that like you're supposed to like make the connectionwith.

So yeah, it's another cool thing aboutour synthetic population, like decision makers. , like more easily connect withthe data that, that they're being exposed to.

Julian: Yeah. And in thinkingabout it from, from a different pers or not a different perspective, but froman outsider's perspective, looking in, one of the use cases obviously like yousaid, is like, public health and how virus is spread and things along thatnature.

And another one is just. Being able tohave the opportunity to, to experiment with so many different range ofpossibilities. And I guess more commonly outside of say, public health in thatsector, what are you seeing the technology really being used for and the usecases it's being used for now and is yeah.

I'll start with that question. I gotanother one for you.

John: Sure, sure. So, oneexample would be in the insurance industry. Mm-hmm. . So in insurance, they'realways trying to predict or forecast what's gonna happen over the next. andall, all the different things that could happen. Yeah. What are, what's, what'sthe range?

And then from that range it's like, howdo you get more certain about it? So, the insurance market big, like, reallygood market for us. The other, so if you think of what we all, what the worldexperienced with Covid, like, let's say Covid was I don't know if Pokemon cardslike were this expansive, but like, let's say Pokemon cards like just becamelike the next new thing.

Like no one's ever seen 'em before. orthere's like a movie that came out recently related to like, what if theBeatles didn't exist or whatever, and creates like that. All right. So let'ssay like the spread of that catching on is really a new product in a newmarket. Mm-hmm. . And so if you're a company taking a new product to marketunderstanding, are you gonna get like some non-linear viral spread of, of thatproduct?

like truly going back to like thetipping point, Malcolm Gladwell example. Yeah. Like I'm launching a product, isit gonna take off? How, how can I change the conditions around like the usercommunity or the customers that I'm trying to get Yeah. To make it more likelythat it'll take off. So marketing and forecasting in consumer goods companiesare another, another area.

Like in the insurance space, we're kindof changing the way that actuaries get to work with data by using a syntheticpopulation in the consumer package goods, demand forecasting, strategy,planning, yeah. Group that exists within big companies. By starting with asynthetic population, we're also. Giving them a whole new way of looking atwhat they get to do with their jobs. So.

Julian: Yeah. And, and oneobviously the skeptic in, in me is like, what happens when it goes wrong? Whathappens when there's a bad actor testing so many different types of things andyou find that they're testing something in particular, and what, what would gowrong in that process? Is it the input of data?

Is it the intention behind that input?I, I can see there's so much ambiguity. Without, I guess, keeping things withina certain degree of, say, requirements or, or certain value set. Yeah. Whathave you seen kind of in your experience or in, in your kind of forecasting ofways somebody could use this in, in, in a say, Malison way?

John: Yeah. So this issomething that I think founders of most technology companies have to grapplewith. It's like, , what is the, the alternative use of this type of a tool?Yeah. And one of the ways that we're like setting up our platform so that wecan combat that is because you're starting with a synthetic population.

Like that's a single source of truth forthe data. So let's say that, like, you can like go back and say, all right,that's where the, the baseline data set came. . And then as you build yourmodel and like there's different assumptions, you could say, okay, this dataset was generated by this model. And the cool thing about what we get to do isthat we're opening, like open sourcing our, all of the models that aredeveloped.

Yeah. Like companies can like say like,we don't want arts to be open. Like, that's fine. It's, it's their project todo so. But our objective is to make the models open so that the assumptions canbe tested. It's known and. , it's like visible for the users and the decisionmakers. Why that's important is the decisions that people use the software likeours to do.

They do impact entire population. Yeah.So like, let's say a mayor of a city running simulations understand like wherehe's gonna make he or she's gonna make investments, like that's gonna impact,know Yeah. Large groups of people across the city. The insurance company, likethey're gonna like make a policy that might exclude a whole demographic group.

Like Yeah. All of these things, like ifthey're. Give individual people an ability to better influence the decisionsand impact them. So yeah, if we use that as a baseline where a bad actor coulduse this data, like let's frame it in the idea of fake news or all of like thelarge language models that could be used to generate a whole bunch of likereally negative content.

Sure. Or like the video image relatedstuff. Like you do the same thing with what ours would be used for is like togenerate a data set Yeah. That could be used. show something that's gonnaimpact a community that might not be totally true or bad. Yeah. Yeah. So likethose types of things would be how a bad actor would use Yeah.

Software. But the one thing about, it'slike you have to look at how much, how much does data influence the behaviorsand decisions of individuals. Yeah. And I think that's like the, the power of atool like ours. Like by making the models open and like one single source oftruth for the data people can trust data more and like can say like, oh, we'regonna move to data informed decision making.

The time where that gets like, turnedaround and use negatively. Mm-hmm. , I mean, that's gonna undermine the entireuse of data and AI to help inform decision making. So, yeah. Certainlysomething that we have to be aware of. Care over time.  

Julian: Yeah. We've seen I'veseen I think, the only company that does synthetic populations in, in, in testingin those environments.

But I've had other companies who kind ofuse this synthetic data to, to, whether it's give insights or come up withdifferent um, outcomes. If, if they have certain inputs in particular, like aproduct or a feature set or something like that is this whole synthetic data.Is it a fad? Is it something that's gonna move away as da?

As data's becoming more and more betterindexed, better labeled and, and better retrieved and, and ho honestly bettermanipulated is this something that's gonna go away? Is there something that'sgoing to say replace it in the near future? Or is it an underutilized toolthat, that will see more of a gaining popularity of it moving forward?

John: Yeah, I think I thinkit'll gain in popular. The other thing to note is like there have beencompanies that have been using what's been rebranded as synthetic data for along time. Yeah. And so I think like a lot of technologies sometimes, likemaybe there's like a 20 year life cycle of like, where one term is like reallycatchy and so businesses use that to like get adoption.

I think synthetic data has a little bitof that. There's like the GaN side of things, like people have been using thatfor a while. People have been, yeah. Replicating data. De-identifying data setsfor privacy reasons for like, a number of years. So I think one synthetic datahas been around.

But I think for those that are new tolike this concept, it'll just continue to get adopted. Yeah. One example I cantalk through is like, a lot of times these, like, it's more like hottechnologies can only get adopted by the biggest companies. Yeah. One examplemight be like Twitter for example, so Twitter, this guy's no longer at Twitter,but I met with one of the guys that ran Twitter simulation engine.

And so what Twitter was simulating wasevery single Twitter handle was represented as an agent. Mm-hmm. . And thenwhat they were running simulations on was any change in marketing or any changein the product. How does that lead to all of those Twitter handles beingmonthly active users? Yeah. And that's a different state in that agent-basedmodel, and so they would just test every decision.

and they would forecast like, oh, likewith these decisions that we make monthly active users goes up. Yeah, it goesdown by this number, whatever. And so being that, that's what their businessmodel was pretty much based on from like a value perspective they were usingsimulation to help answer those types of questions.

So if we say like Twitter and the next,hundred largest companies around that size are using simulation, how can webegin enabling. Yeah, everyday people or smaller mid-size businesses to accessthat type of technology. And I think by having our synthetic population outthere for people opening the models so that like others can just like use themas they want, those are all things that promote like an even larger adoption ofthe technique.  

Julian: Yeah. Yeah. It's sointeresting, especially the kinda the open co-op, open source play that, that alot of companies either will be fully behind or not. And it sounds like yearone. In, in, in the, in the boat of sharing and sharing that information, doesthat also help train the models to be more accurate or just create morepossibilities that maybe you, you hadn't predicted or other companies hadn'tthought about?

That overall kind of enriches the, thedataset.  

Yeah.

John: There's probably likeways to view that as like enriching a data set, but also just enrichingunderstanding of how the world works. Yeah. So, yeah. Inherent in our nameEpistemix, the stems is epistemology. So like the study and understanding ofknowledge creation.

Yeah. So like that's something that likewe look at as being able to better understand or generate knowledge. Is likejust a overall good thing for humanity. So, yeah. Yeah. Like we hope that byhaving the models being open, that it can inspire, creativity Yeah. To likesay, oh, this model's applied here.

I'm applied in another use case. So oneexample of this would be oh, let's see. We have a, a membership model. Mm-hmm.. So like, let's say you're, an organization and the value for yourorganization increases with the more members that you have. Mm-hmm. ? Well,that same membership model, if you pick it up and apply it into a politicalcontext, it's like, oh, cool.

Like, we're trying to get more people tovote for our candidates. Sure. We're, we're a credit union. We're trying to getmore people to sign up for our credit union, or I'm an individual influencerand I'm trying to get more people like engaged in my content. It's a, it's amembership adoption or membership acquisition model that people can pick up anduse in all sorts of different ways.

And so, a lot of human behavior, youcould say is similar, but like, when you change the context, you're justchanging the parameters of what's influencing it. Yeah. So that's that'ssomething that we believe by open sourcing can lead to larger adoption. We'realso like pretty plugged in with different academics.

Mm-hmm. and like the research c. whereopen source is really important and if we're going to like, have the name andlike tout that we're all about human understanding and like truth, like we needto make, if we were closed off in the models, like that doesn't align with likethe values of who we are as people, so,  

Julian: Yeah. I love that.And, and tell us a little bit about the traction. How many people or how manycompanies are using Epistemix now and, and what are you particularly excitedabout this next phase of growth this year and beyond?  

John: Yeah. Right now, 153.Nice . And so we're, we're excited about that.

It's across like a number of differentindustries. I'd say like the next phase of growth is really around likenarrowing in on product market fit Sure. And how our software can be appliedacross the different verticals where people are interested in using agent-basedsimulation or working with a synthetic popul.

I hired my first salesperson a monthago, so we're now at like the point of like adding sales and adding customersuccess. Right. And culturally, how does that like, fit in with an organizationwho's been very r and d and engineering focused? Right. Yeah, so for like theengineers, productory people who are considering starting a company, it's like,I guess you can kind of start on two ends of it.

You're. Building a product like for likea group of users. Mm-hmm. , and then need to figure out, okay, how do wecommunicate this from a marketing, sales, branding perspective? Yeah. Or you'restarting at like the have this idea and I'm gonna market and brand and likefigure it out and then like build the company backwards from, from there.

Sure. Yeah. Yeah. There's all sorts ofdifferent ways that people come about it. We came about it from the like reallydeep science side of things. Yeah. are now moving towards being a more customersuccess oriented organization.  

Julian: Yeah. Yeah. And whatare some of the biggest challenges that you face today?

John: So like any founder ina, like, we haven't done a Series A yet. Like we're ready to do that prettysoon. So let's say like any founder that's raised Seed Money series A. , you'realways worried about your runway. So, so current funding environment has beenreally tough over the last, let's go 10 months or so.

Mm-hmm. . And then like the event lastweek with svb, it's, adds like another wrench into like any deal that wastrying to get done this week. Mm-hmm. like that certainly delayed. Causewhether that fund has their money or like they were uncertain about their SVBthing, if they're switching banks.

They're, they're likely to have had aportfolio company that was having to deal with some, some panic. So, there'salways like the runway and funding type of challenge. There's always the makingsure the team is the right team, still believing in what we're doing. Like,yeah, running in the right direction, running the same direction is, almostmore important than Ron in the right direction.

Right. So. Yeah, I'd say there's,there's always the funding connecting with customers, having the time to likereally get the messaging down right for the people that you're hoping to serveand work with. And then, you always have to be concerned about the culturewithin your team. Yeah.

Yeah. So those are always the thingsthat keep you up.  

Julian: Yeah. Yeah. Ifeverything goes well, what's the long-term vision for the Epistemix?  

John: So long-term visionis, Across in the introduction, there's the health modeling, the socialbehavioral modeling the environmental side of things, the economic models thatour synthetic population and using the synthetic population with the modelinglanguage is just used across all of these different disciplines.

And what that enables is anunderstanding of people and having a conversation because it's same datasource, right? Which like provides a single source of. The models enable you tobe open about your assumptions, and that creates more dialogue. So then you canhave like, just like more open democracy, more open decision making especiallyon things that are influencing entire populations of people.

So our goal is that we kind of becomethe standard for HM, based simulation. And a simulation grows in its use casefor helping decision makers with high stakes decisions that impact populations.So yeah, ideally, . Yeah. Tens of thousands of users around the world. Yeah.Yeah. That, that's where we wanna  

Julian: go.

Amazing. And I always like this nextquestion, I call it my founder faq. So I'm gonna hit you with some rapid firequestions and then we'll see where we get. Sure. So, first is what'sparticularly hard about your job?

John: Managing my ownenergy. Yeah.  

Julian: Yeah. And in, inregards to, if we think about similar technologies that's out there and, andwhat associates with agent based modeling how is it different from AI and howdoes it, or is AI incorporated it? What's its relationship to some of the moreintelligent, software technology or.

I just products out there that we see,how is it different, but how is it similar in relationship to, to AI and ML andthings along that nature.  

John: Yeah. So some peoplewould argue that like AI's within simulation of simulations and ai. Yeah. Youknow what? I'm indifference to, to that like, you people are gonna like, havetheir religious doctrine about it.

Yeah. I don't, I don't need to commit'em if they're already, a believer in one or the other. Yeah. I would say thedifference between H based simulation and machine learning is machine learning,getting more and more narrow. And sometimes those models are not explainableentirely. It's like the math all works and like can get the understandingthere, but like, yeah.

Alright. Is that like really theoutcome? Don't know. We're supposed to agent based simulation, you can go backand look at every single time step, every single state that your populationthat you're simulating. Yeah. What, what state they were in. So the idea ofexplainable models I think is.

Difference, but also is like how AIagent based simulation and machine learning can complement one another.  

Julian: Yeah. What are somethings that you've seen? What's one, like really interest, interesting andfascinating use case you've seen the product being used for, and what's onethat you haven't seen it being used for, but that you would like it to see?Like, like to see it used for.  

John: Sure. So, onefascinating one. There's the think tank for the broadband. , the challenge thatthey had was like understanding what is gonna be the adoption of like new athome, like VR type headset type things that is gonna put an increased strain onbroadband requirements.

Yeah. And it's like they're trying toforecast consumer tech adoption to then plan for the future of what thebroadband network needs to look like. And they use like our platform to startsimulating that, which I think was pretty cool. Yeah. . One example of wherelike we have a, we don't have a ton of traction, but would love to see thetraction grow is in health equity officers.

Mm-hmm. in health systems or likegovernment related with the whole idea of you have pots of money that you'regoing to make investments in. Yeah. That ideally should improve a healthoutcome for a population. And by simulating all the interactions, you're openabout your assumption. And ideally the money that's spent can actually improvethose health outcomes over time.

Beyond that, I'd say like education inthe United States, it's like, we're, we're looking at like, okay, we'reinvesting all of this money in education, but like, are we really getting likeimproved outcomes? Yeah. And if we're looking at improving graduation rates orthe eighth grade reading level, like, can we get more people able to read atthat level across the country?

Right. Like what compounding effectsmight that have in other. It's like those are the, the cool use cases that I'dlike to see people start to use simulation for.

Julian: Yeah. Yeah. It's sofascinating the, the different directions that this could go. And I'm eventhinking about myself, different experiments I wanted to run, but it hadn't hada population to, to test them on or I haven't had the, it takes a lot of time.

Right. And, and to be, the ability tohave something do it almost on the back end while, while it's running and thenhave this interface is extremely, I think, super valuable. Not only for anindividual, but as you said for. What's something that as a founder, you knownow wish you learned early in your career?

John: I learned somethinglike that every, every day, every week. I would say the importance of beingvulnerable in the team. Yeah. I think that builds a lot of credibility and itinstills trust. Mm-hmm. , so I, I went through this with. , some of myadvisors. It's like, John, like, you don't, if you're not open about what thechallenges really are, we don't know and we can't help.

I'm like, that's why we're here. And so,I've, I do encourage that behavior with my own team. Yeah. And I to do it bymodeling it myself.  

Julian: Yeah. Yeah. Well,that's a big one. I think that's, that's probably one that's not said enough onthis episode or on this show, which is the credibility piece that you said, thecredibility that you gain from your investors, from your team, and probablyeven for yourself, I'm sure, in, in the ability to not maybe be so, whetherthey're tied to the outcomes or however, or feeling that you have to fix everyproblem, but really leaning on the network of people, and I always like to askthis as well, is where do you find that network of, of founders, investors?

Where have you gone to really build ateam around you or continue to go to build? That that network that supports, orat least is, is they're with you along the journey.  

John: Yeah. Like it, ithappens organically. I, I haven't totally like, pinned down exactly how I'vedone it, but yeah, it's been something I've been doing since I was a kid.

Yeah. Even when I was like doing sportsor, asking, other like grown up questions, I've, I've kind of always had a, agroup of older people who have gone through what I've, what I'm setting out todo. and I can learn from their mistakes ideally. Yeah, that's kind of how I, I,I learned best from when I mess up, when I have a success.

I might not, look into it as, or be ascritical about it, but Right. I think when you, when you're really in it as afounder and you've, you talked to others who have mm-hmm. been there before,like you want to look back and, and help others. So I'd say, it's usually likethrough a connection of an existing advisor who's like, oh, like I haven't gonethrough this, but I know somebody who has, like, you just make the intro.

And I think that's, those are the, thestrength of those like loose ties and like people that you, the more open andhonesty are with them, like the more they'll open up their network for you andsay, oh, like so and so also went through this chat with them and it kind of justbuilds over time.

Julian: Yeah. Yeah. It's sofascinating to think about, that it does build so organically, but it's, it'sthe, it's the question, the communication that in the conversation pieces that,that continue, that growth, and it's a lot of activity, but it's much small,consistent activity over time. That, that leads to that, that growth.

It's not just like, Hey, here's this oneperson, and then all of a sudden you, you meet everyone. Right? Sometimes it'sthat case, but oftentimes it's, it's. I always like to ask this questionbecause founders are so brilliant at extracting knowledge from anything thatthey ingest. Whether it's earlier in your career or now what books or peoplehave influenced you the most?

John: The book that reallyset me on this path is called the Health Gap. It's written by Michael Marmet. Iwas, it was the book that inspired me to go into public health that gave me thelanguage to describe how I viewed the world. Yeah. And. Yeah, like I, they eventook it to the extent of like, this guy Len sign at at Berkeley.

He was the guy who taught Michael Mart,like his understanding of social epidemiology and social determinants ofhealth. And I even applied to like, do a PhD program Yeah. Training under thatguy. Like that's how, like how much of like, how into that I was, so yeah,health gap big one. And then, The the Natural History of Innovation was anotherone by Steven Jonathan.

It's kinda like a, a more popular book.Yeah, tho those two are, are really good. And then for any, any founder who'sbeen in it for a while the hard thing about hard things is yeah, probably thebest.  

Julian: Yeah. Yeah. Yeah, it'salmost, it's almost gospel. Now for a lot of founders and, and it really lendsa lot.

But I love the other ones in, in howkind of that, that brought you into the journey you are now. And I guessanother question that I didn't ask that I, that I meant to ask a little earliercame back to me as a founder I'm curious in terms of like your business model,how did you define. Your pricing or how you, how you, your relationship withyour customers, being that, the software sounds like it could be, if, if, ifour on your end it could be sold by how much usage, how much, how much data areyou utilizing, how many models are you training, or, or how many outcomes areyou looking for?

What variables are you looking at? It?It could be subscription model, it could be mm-hmm. based on something. Wheredid you land? Or where did you land currently? And has that always been thebusiness model you, you set forth?  

John: No it, it's evolvedover time and the, the best way that I could like frame it is you have to priceto what your customers find their initial value in.

Mm-hmm. . Because if they're like, oh,like, well, I didn't find this thing valuable, but like it cost me such andsuch amount. It's like people don. Don't connect with that. Yeah. So weactually price to the synthetic population now. Wow. And so, because like,people like, like see immediate value in that.

Yeah. And so we framed, we had toreframe like, oh, we're selling this simulation engine. Well, people are like,okay. Like, well, what do I do with that? Oh, the synthetic population. Yeah.And all these models and they're like, oh, these are like the useful featuresof this thing. Yeah. Rather than saying, , you get immediate value.

Like people get value out of data andwhen they can turn data into insights, yeah. To then make a decision or inform astrategy. Like that's, those are like the valuable things that people latchonto. So we adjusted our pricing to align with our synthetic population becausethe way that we've put that data set together, people can get, immediate valueout of it and anything, any future. is also built on that synthetic population.So yeah. Yeah.  

Julian: Yeah, yeah. It'sfascinating and, and it's fascinating how those things changed. And I love thatlittle anecdotal message there, which is price to what your customers findvaluable, because I think that is something that is, goes under undervalued orunderutilized in a lot of companies as they, look at popular pricing models or,or ways that people are interacting, the different features that you're using.

But boiling it down to that something sosimple I think is, is something that, that a lot of. Can, can, can use andextract. And I always like to ask this question because before we, before weget into your cls, I know we're close to the end of the episode here. Before weget into where we can find you and how we can support you, is there anyquestion I didn't ask you that I should have or that you would have liked toanswer?

John: The question that Ilove that you did ask is like, dream use case that is not happening already.Yeah. So like, I, I'm glad that you asked that and like, . Yes. The companyyou, internally we use this phrase we're a for-profit with the heart of anon-profit. Like we want our software to be used for good.

Yeah. And so that's something that, thepublic health use cases, like want drive on that, the equity use cases wannadrive on that. The improving decisions that are happening at the citygovernment. Like seeing all those things happen would be a dream country forlike founders and like the people that join the team.

Julian: So, Yeah. Yeah, yeah.It's incredible to see not only how your technology works, but also how it'sutilizing the different use cases. And I'm so excited to share this with theaudiences and for any founders out there who, who want to start testing and,and utilizing the platform, where can we find you? Where can we support, wherecan we find Epistemix and where can we support you as a founder?

Give us your LinkedIns, your, yourTwitters, your websites, anything that we can, we can use to connect withyou.  

John: Yeah. Atepistemix.com/platform is how you. We got our fully automated, go to thewebsite, log in, start building models on your own. System set up on December7th. That took about two and a half years to get to.

So like, yeah, that, that's the, theplace to get access to the software. The next conference I'll be at is Vive inNashville. It's a healthcare focused conference. LinkedIn, just LinkedIn, JohnCordier. And yeah, interested from a customer perspective, partnershipperspective, investor perspective. Glad to, glad to connect.  

Julian: Amazing. John. It wasso interesting learning about your background, your experience, but alsoEpistemix, how the technology works, what it is being used for, and what futureways it could be used for as well. And and also all the little snippets ofadvice and wisdom that you've gained as a founder has always been super helpfulfor the show.

And I hope you enjoyed yourself andthank you again for being on Behind company Lines today.  

John: Yeah, thanks. You madeit easy. Appreciate it.  

Julian: Cool. Yeah.

‍