eMalick Njie, Ph.D | Part 3: Shaping the Future of Genetics Research with AI

Intro - 00:00:01: Welcome to the Biotech Startups Podcast by Excedr. Join us as we speak with first-time founders, serial entrepreneurs, and experienced investors about the challenges and triumphs of running a biotech startup from pre-seed to IPO with your host, Jon. In our last episode, we spoke with Emalick Njie about his move from Boston to Florida, the cultural differences he experienced, and how the move revitalized his passion for research. We also heard about his PhD work at the University of Florida, his focus on the cellular and genetic causes of Alzheimer's disease, and his insights into the critical role of microglial cells and his experiments with stem cells and plaque-busting proteins. If you missed it, be sure to go back and give Part 2 a listen. In Part 3, we talk with eMalick about founding Genetic Intelligence and its mission to combine AI and genetics to tackle diseases. He recalls the challenges of founding a company, the personal sacrifices he made, and the eventual breakthroughs that earned recognition from the National Science Foundation. eMalick also reflects on how company culture and leadership shifts led him to partway through Genetic Intelligence, sharing the lessons he learned from his journey and the importance of building teams aligned with the original vision. 

Jon - 00:01:30: Okay, so now you're like, okay, I can't close this door anymore. And you're founding Genetic Intelligence. Can you talk a little bit about Genetic Intelligence mission and focus and what were the early days like for you as the founding CEO?

eMalick - 00:01:45: Yeah, absolutely. So remember the whole thing about name and genes is very sort of utilitarian. Like you got to do that as a Genetic Intelligence. Let's put this together. 

Jon - 00:01:56: There we go. Very C-elegance. Very C-elegance of you.

eMalick - 00:02:01: It's so utilitarian. And we've got the second name that ages well in retrospect. So, I mean, starting Genetic Intelligence, I would say it was the hardest thing I've ever done. I buy an exponent. I don't think I'm like super great in the lab, but I'm set pretty well to lead a future lab as a professor studying in Sheetman. These genes are so many, I could study them for the rest of my life. And I was taking a hard turn to explore and what we could do with AI and genetics. Without much support. So AI was something that was not popular at the time. It was actually considered like a, there's multiple AI winters there, so it was considered like a graveyard. Luckily, I didn't notice.

Jon - 00:02:49: Yeah. Sometimes that's all you need. 

eMalick - 00:02:55: Yeah, I didn't notice. But, you know, I remember going into Columbia Technology Ventures office with this idea for a patent. And I was so nervous walking into that office. And now I realize that. I walked into that office with the biggest bag of gold I have probably ever walked into that office. But I was so nervous and... Like there was not confidence. From what I was doing. I was just sort of naturally trying to pretend I'm confident. And then, you know, moving, like telling Marty that I'm leaving the lab is one of the hardest things I've ever done. And we still chat about it now. I spoke with him in person last summer. I think we're still going through the trauma of me, you know, myself, to be honest, because it was actually such a wonderful experience. And I also had no money. I literally had no money. So I took out my 401k. The retirement plan. And that gave me about $10,000 because they punish you heavily for withdrawing it before you're 65 years old. And... I basically funded it by eating less. I'm not even joking. So I went into this. So Move 300 is a good reference again. I went into this space where I was waking up at 4 AM. Every morning going for jogs, which I'm doing now again. But back then it was new to me because I'm not a morning person. I had to shift that. And I'll go to the supermarket. I go to the vegetarian vegetable aisles and I'll look for. Like broccoli and all the lettuces and so forth. And I look at the weight. So price per pound or price, you know, for LB. And just whatever's the cheapest thing is called lunch and dinner, man.

Jon - 00:04:46: Yep. By the way, again, you're not alone in this. So Exceder was the same way. So we're a weird company because I was a bench scientist and I didn't know better. But we haven't taken outside money ever. And so all I knew was that, like, we need to. Bring value to the customers so I can earn the right to pay myself. Obviously I was talking about it earlier when you have zero customers, there's no money to pay yourself. And so we would do the same thing like early days. I remember going into Safeway and really scouring what is the cheapest thing that we can, we can stretch here. It was the frozen aisle.

eMalick - 00:05:26: Ah, nice.

Jon - 00:05:28: The frozen aisle, frozen vegetables. So like, look, I wasn't completely just like going off the rails, those, but. Like, and we were able to make like, really, it's like knowing what rock bottom is. And, and, and again, I'm speaking for myself now, but like. Knowing and this is like a long time ago but, like I still remember it's searing me today, like figuring out what rock bottom was and I like it was hard but still had a blast. And Part of that that's incredible is that when we're thinking about what is the bar to happiness, It's not much. Like it's not much because I was like, I was trying to make, you know, I was trying to make 15 bucks a day stretch. So, and me and my wife were still having a blast and join her like each other and building Excedr. So I know exactly that feeling where the first piece of equipment that we bought. Was from my life savings, like exactly, like tapping 401k. So, and that was just one instrument for one client. It was a flash chromatograph. I was like, I'm doing this. It was like a medicinal chemistry lab. And I was like. This, this could have like this this is all, like you know all in, and but again, I didn't know better kind of like, exactly what you're talking about. 

eMalick - 00:06:48: Yeah, yeah, yeah.

Jon - 00:06:48: I didn't know better, I was like it's gonna work out.

eMalick - 00:06:51: In a sense to that right.

Jon - 00:06:53: It's going to work out. And, you know, it could have very much not worked out. Like, it could have been very, very bad. And that's kind of like entrepreneurship, really. And so, sorry, I didn't mean to Bogart your story there.

eMalick - 00:07:06: No, this is actually what helps me move along every day. Like, you sharing that experience with me, at a Safeway, learning, like, hacking the Safeway to find that, like, or exploring the Safeway to find a hack of the frozen isle. Like, we are kindred spirits in that, and that literally motivates.

Jon - 00:07:24: Yeah, it was frozen vegetables, cans of beans for protein and fiber, and, like, a protein if we can afford it. That was years like years of doing that.

eMalick - 00:07:35: Yeah.

‍

Jon - 00:07:35: And so those are, those are the, you know, those are the roots and the kind of like foundational things, I still don't forget it at all. 

eMalick - 00:07:42: It's so important. And it teaches you to be, one, to appreciate everything that's come subsequently, but two, to always be capable of being lean, which we're going to get into Ecotone. I'll share some of those lessons about that leanness from that period that allows to be like a very, no saturated fats operation.

Jon - 00:08:02: Yes. 

eMalick - 00:08:04: It's a mess of values created. But yeah, so I guess by being very frugal and... And sort of this dogged determination, that we have this thing called a universal approximator. Like my patents that I wrote, it's like in the first lines, but it ended up being two of them. The first lines are like universal approximator, universal approximator. And it's so true, right? Which is why every industry is being impacted by AI today. So the vision of Genetic Intelligence was to explore the use of artificial intelligence within genetics, right? And would impact, of course, to different diseases. So I shined a light a lot on Alzheimer's disease, but again, universal approximator was pasted everywhere, right? And what we did was start putting, following up on what I was doing in laboratory, start putting now human genomes, like a C. Elegans genomes, but human genomes in convolutional neural nets. Can convolutional neural nets accept genetic data? That was an open question, right? Can they take some of it? Can they take all of it? What can you do if any of these things occur? So I would say one of the most wonderful things about genetic intelligence is I believe to this, at least to my knowledge, that we are the first to put an entire human whole genome within a convolutional neural net. This is very forward thinking. And we got rewarded with the National Science Foundation giving us an award for...

Jon - 00:09:40: There we go.

eMalick - 00:09:41: The ideation of this and execution of it. And so, yeah, money started to pour, sort of slowly trickle in and pour in, and eventually like several million dollars in the bank. And some absurd, to me at the time, it was like absurd valuation. But it was just hard work. Every step, convincing people of artificial intelligence was harder than convincing them of so many other things. Which is so like now, you know, I don't have to even do that.

Jon - 00:10:15: Yeah, now you're just like, like now it's like being pulled from you, but I can't imagine like, this is like pre-ChatGPT, like where no one has any idea.

eMalick - 00:10:24: Exactly. But it was like a momentum that I saw. Like there's like, so these larger shifts that occur, like think of like sort of like baseline things, like going to a stadium, like a busy stadium, there's like a movement of people in it, right? So if you just zoom out, it's like this big shift of movement in, right? And no one individual responsible, it's a collection. So even if you're like a laggard, you eventually will, it's like a movement. And then at the end of the game, it's like a movement out. It's a larger movement that's happening. So even if I'm telling you that there's going to be a movement out and you're like, there's no movement out, it's just a matter of time before I see you.

Jon - 00:11:05: You are going to experience it. Yeah, you're going to experience it.

eMalick - 00:11:10: Exactly. You're going to experience it. So I just knew it was just time. And we luckily had enough early investors that saw this movement for us to be able to chug along and do some interesting things. And then, of course, now we've seen that movement really, really take hold today.

Jon - 00:11:28: Very interesting. And I'm thinking about that experience that you mentioned, like going to the Columbia like patent slash tech transfer. And was it just you or did you have like other conversations? With like comrades and co-founders who are like, let's do this thing. Did you did you kick this whole genetic intelligence off just Solo-Dolo?

eMalick - 00:11:48: Yeah, at the first year and a half, it was just me. I mean, I'll be honest, it was very lonely. Yeah. I remember going to the tech ventures office by myself. I remember going to the little library, which is Columbia has like a, just people that like stamp, give you an official stamp that, you know, so I could send the company to the government for doing business as, or something like that. And yeah, it was just like a vision and I was just sharing it with people, but I didn't know any venture capitalists. Columbia isn't much better now, but at the time, they didn't have the same spirit as Stanford. So I went to the computer science department multiple times to talk to people about AI and they really weren't into it then.

Jon - 00:12:34: So again, you're not alone. I was at a dining room table. I didn't know any investors at that time, which is part of the reason why I was like, well, if an investor... Is not gonna pay this salary. I need to find it somewhere else. And so like bring value to customers. But I think that's like something too I'd like to underline is that. You might hear about entrepreneurship in the media. It's like this like glitz glam kind of like just super just like exuberant kind of like. Thing, but I think a lot of the time it is lonely. And especially those early days. And I was for me. It was, I've never done sales before. So, but I was told it was like, you just gotta. Pick up the phone. So tried my hand at like years of just cold calling, just cold calling, trying to garner some business. And that first piece of equipment was off a cold call. Um, and the timing was perfect, but it took like, it is brutal. It is super brutal. And, but again, I think we're all stronger for it. Like those it's kind of like going monk mode or like hermit mode where, or I, or if we, you know, go to Dragon Ball Z, it's like the hyperbolic time chamber. You increase the gravity, like 10, 100 X, and you're just like doing pushups. That's me cold calling, but in Dragon Ball Z, you're just doing one handed pushups. 

eMalick - 00:14:01: Exactly.

Jon - 00:14:01: It seems like an eternity. And then you leave and you're like, Oh, ChatGPT. For me, it's like, Oh, people are starting to get it. And from our perspective, you're, you're, you're at the frontier, right? Like AI is like. Very much more frontier than what we're doing. Like, right. 

eMalick - 00:14:18: No, but you hit hard and nail. I like the Dragon Z analogy. I told you how it converted from being a late-night person to waking up at 4 a.m. Like, Dragon Z, man.

Jon - 00:14:27: Yeah, yeah, exactly. Adult swim in Toonami. Exactly. And so, for me, like, what was super confusing for me is Exceder. So, we provide lab equipment leasing. Leasing as a concept. It's an age old industry. Like everyone's like leased an apartment at one point in time, probably, you know, probably there's a majority. So people are familiar with it in that regard. And other industries, you know, we're talking about like applicable to other industries and business models. Leasing is used like commercial real estate, leasing a car. It's everywhere. It's pervasive and it's been here since the dawn of time. So when I first started Exceder, and still to this day, by the way, there's more knowledge around it now. But people are like, what the heck is leasing? I'm like, what the heck is leasing? Hold on, wait a second, have you rented an apartment before? Like maybe, or is there something, or maybe a friend has rented an apartment before. And, but for you, I can kind of empathize with the kind of just like trying. And I, so kind of exactly what you said, like there is this motion. If you zoom out. People are going to experience it eventually. I just don't know when. I don't know when the game, this is where maybe the proverbial game example kind of ends here. But I don't know when this movement is going to happen, but I know it will. Because I've seen it, how it can be beneficial elsewhere. And so I empathize with that.

eMalick - 00:15:53: Also, the thing you're saying with Elisa, and it's sort of like a paradigm shift in people's mentality. And they need some time for some cognitive dissonance to resolve itself.

Jon - 00:16:02: Yes.

eMalick - 00:16:02: And that's called time.

Jon - 00:16:06: Yeah, that's exactly it. That's exactly it. It's time because I'm like. I'm trying my best to explain it, but sometimes you just need time to, for you to, or the community or whatever, this movement to catch hold. 

eMalick - 00:16:20: So you literally give immediate value, but it takes them time to recognize that value. It's like a latent period before they recognize it. And they're like, why didn't I do that before? 

Jon - 00:16:30: Yes, yes, that's exactly it. And I always think about kind of like that normal curve distribution where you have like early adopters, like middle adopters and late adopters. And, you know, I think there are, I continue to have conversations with people who are the early adopters and they're like, they just like quickly got it. But now we're kind of in this realm where we're in the kind of the bulk of the normal curve where people are like, I wish I did that during my first venture, but like, here's another go, second venture, let's do it this way. And you kind of learn. And typically it's via like lived experience where you're like. I could have done it a different way and I want another shot at it. So I'm sorry, I'm now derailing us a little bit here. So now, early days, it was you. It was hard, but you then started to get the grant funding and got investor attention. Is that when the team started to grow and you started to kind of, I mean, I'm going to imagine you're a founding CEO. It's not just the technical component of this. There's the whole company building and everything to that. Can you talk about that? 

eMalick - 00:17:35: Exactly. So, yeah, you have to build a company around the vision. And so with enough time, I was able to get the interest of a good friend of mine that was also at Columbia with me. His name is Patron Adave. So Bertrand was in the chemistry department, which is, I would say, a three minute walk from my building in biological sciences. And we had known each other and he was he had finished up. His PhD, and then went on to McKinsey and Company as a consultant, which the combination of One, having a PhD in one of the hard sciences, and two, having business experience was something that I valued. So I told him about this crazy idea. Literally over a napkin.

Jon - 00:18:26: Very startup, very startup, Yeah.

eMalick - 00:18:30: Exactly, yeah. And I was like, so I want you to leave your high paying fancy job, and come slug it out with me, and this is going to change your life um think about it.

Jon - 00:18:41: He's like, what?

eMalick - 00:18:43: So he did. He went and plugged in all the numbers using McKinsey software and all that stuff. 

Jon - 00:18:48: Yeah, yeah, yeah. As they do.

eMalick - 00:18:50: He saw that this was something up the horizon. It wasn't just my word. He went and did the research. And he came on. At first as like a business assistant. And then eventually things were really moving along. We had gotten some more people. And I wanted the development cycle to just accelerate. Put an exponent on it as much as possible. So he was like, have him become the CEO. And then I could focus more on the code base and the bills. Which worked out for what we were doing. We were able to attract some more investment. And then I was able to lead us into getting it. You know, I mean, so the, um, into the carbonate and just build different versions of carbonates to explore what they could do. And the team just started to grow, grow and grow, which has positives and negatives, which if you want to, I could get into on that.

Jon - 00:19:48:

I am interested because I think a lot of companies are at that inflection point.

eMalick - 00:19:52: You know, oftentimes we are propelled by what you're supposed to look like versus what's best for your business. So having like a big office is something that people look at as like an indicator of success. Having a lot of people is like an indicator of success. Having origins such as like a car garage, like a Google founders is an indicator of success. But you have to remember before those guys. Their business out of their car garage that was looked at as like an indicator of fail.

Jon - 00:20:23: Mm-hmm.

eMalick - 00:20:24: Right.

Jon - 00:20:24: Yup.

eMalick - 00:20:25: So we were innovative in a sense that we just had a company be completely from home. This was 2016, so way before the pandemic. And that was looked as like an indicator of fail. You know, what are you going to do? But we stuck to our guns.

Jon - 00:20:42: Before you move on to the next point, again, you're not alone. During that cold call, like flurry that I was doing. So I talked about the win, but I still am scarred today because I won't name the company and I won't name this institutional investor. They picked up the phone and basically berated me on the phone for being there. You sound like a one person shot. Do you even have an office? And, you know, this is before work from home was cool. Yeah. And so I was getting chewed out for even daring call them for, and not having like a, you know, being in a skyscraper and, and to this day, just like scarred.

eMalick - 00:21:28: Yeah. Yeah. That's such a terrible, like. Thing to do and, yeah.

Jon - 00:21:33: Totally okay, totally like, like you, know you live and you you live and you learn you move forward, right? But it like it definitely, exactly what you said, it wasn't cool, like it wasn't cool, until it became cool, like you know now I'm dating myself, but I'm like Oh, like those like really baggy pants that I used to wear. Like they're now cool, but I used to get made fun of for wearing those baggy pants. Anyways, now I'm dating myself, but sorry, you're not alone in that.

eMalick - 00:22:00: Thank you for sharing. Yeah, it's so true. It's like those individuals that exist for, you know, sort of bullies. They spend less usually starting things on their own and they love to give opinions. We stuck to our guns and that, uh, and certain wasn't cool. Um, a lot of my folks back at Columbia were like, what is it eMalick doing? Like I've talked to them recently and they were like, we thought you were crazy. I was like, I was crazy. 

Jon - 00:22:24: Yeah. No, no. Still crazy. Yeah.

eMalick - 00:22:27: That was called the right crazy that you should have been too. 

Jon - 00:22:29: Yeah. Yeah, exactly. You should have just come along.

eMalick - 00:22:31: Yeah. Yeah. Um, well we used, we built a company. So eventually, you know, eventually we're out on this like 15 people, 20 people, uh, within different levels of like employment status to advisership. You know, I sort of started to lose track of what the heck is going on in my company. Um, so I'm going to just, I'm going to now just quickly go to the down phase of Genetic Intelligence. So while we were busy not being cool, like doing work from office, but still like having this vision of the world, that was the correct vision. Um, we didn't know that we were the first like modern, AI company out of Columbia. There was another company, I think in 2012 or 11. Um, so these companies are based on like a different generation of AI. They're not using neural nets, like back propagation, having neural nets, um, the way we understand them to be today. So we were like the first modern AI. I'm at LinkedIn. I have medical company, but it's actually companies in general. And so we got traction eventually, particularly after the National Science Foundation Award. And I had given Batran the CEO position, which worked for a while, but then didn't work after a while. Long story short, the people that I started to see within the company just didn't share the ethos and the vision that I had set out for the company. Vision being how can we really hit a high note on exploring genetics within artificial intelligence systems um would i can focus on disease to make it relevant today right it became more of a consultant energy that i didn't find to be the most comfortable for me to be um and by the time i was like this is too bad that i had to start changing it it was essentially too late um so listen to entrepreneurs um not only build your team but make sure you always have your team aligned to your vision this is just key um so your hires have to have to always sort of you know do the pinch test to make sure that they are aligned there and if they're not do not be um hesitant in letting change and change in staff as simple as that um this is a boat that you're building it's super important and if it goes the wrong direction you and them will be hurt so it's important for you to have the right people on board when I realized I was essentially outnumbered. Long story short, I had developed some tools that allowed us to look at the heritages of different people for the purpose of being able to separate disease group individuals without having their heritages confused. Downstream process and basically prevent hallucinations down the line in a neural net. But the desire of some other groups within the company to monetize this immediately, to try to become a 23andMe was very strong. And I didn't build genetic intelligence to be a 23andMe competitor. I built it for us to do much different ambitions. And I just couldn't change it. And despite everything that I tried, I just didn't have the strength. I hadn't done a background to basically ensure everybody was on my side the entire time because I was so busy writing the code. So I cut my losses and sold my shares and left the company to start anew. 

Jon - 00:25:56: And so there's a bunch of things that I kind of want to like kind of touch on. So. Your comment about like Culture, being intentional, and really being intentional with your hires is, I think, incredibly important. Because I think sometimes, exactly what you said, a lot of the business media having a fancy building with a fancy address with the highest headcount and the largest funding. Looks great on the surface level, but it can be incredibly nefarious if that's what you're optimizing for, because much like a lot of things, people and cultures are compounding. And you think just one hire that's maybe not exactly a fit is okay. And when I say fit, everyone has a fit, like the company fit. We have Exceder. We have the Exceder way of doing things. I'm not telling anyone that the Exceder culture should be your culture. I'm not saying that. I'm just saying for whatever culture that you are seeking out to build. Like not following your kind of your principle on it. And you're finding that alignment is actually what starts to create this like negative compounding effect where the culture compounds in the wrong direction because that person changes the culture. When the culture changes, you hire your next person, you hire just. A little bit more off, and then a little bit more off for every successive hire. At that point, you don't recognize the company that you originally founded. The problem, I think, when I saw in 2020, 2021, is that everyone was optimizing for those surface level things at great speed. Now, we're dealing with the indigestion of it. Right. I know it's great to make these grand announcements. It feels great. But if you just take a longer view horizon on it, just really think long and hard, what does this mean for five-year? Insert your company, 10 year, insert your company here. And. The people stuff is the hardest to change, too, because people are involved. And it's an emotional decision. And so I think that's really important for anyone who's listening out there. That just think long and hard, just think long and hard about those hires, because it's not just that in isolation.

eMalick - 00:28:20: It's an exponential that destroys your company. It could cause immense strife, or if you do it the right way, you could also build the place you want to be at. 

Jon - 00:28:33: And that's the flip, right? You live by the sword and you die by the sword. Like it's the, if you get it right, that's what super magic happens.

eMalick - 00:28:40: Right. 

Jon - 00:28:41: And. I think we always have that hiring slow rather than fast for us has always been the key. At least for Exceders perspective, and again, we're just a weirdo company. Our idea for hiring at Exceders that... We would love every hire to come in to want to spend their whole career here, like from now till they stop working. And that is super weird. Like it's, and I take great inspiration. I talk about visiting Japan, all this stuff all the time, but the companies that are in Japan, like in Japan has like a disproportionate amount of businesses that are a hundred years old. I think they have like 65% of all hundred plus year old businesses. And it's a weird, it's kind of a weird dynamic, but I take great inspiration in that because when I, when I think about long time horizons like that, it's like, what are the cool things that can be done when you are not worried about whatever it may be? And this is why I like, I don't think we'd ever go public. This is like the, the pressures of this short term, short termism. I'm preaching for John here and I'm preaching.

eMalick - 00:29:49: No, it makes complete sense. Looking at the scene over the horizon is key. Like those initial visions that they found that could last way past your lifetime. 

Jon - 00:29:58: Exactly.

eMalick - 00:29:59: So you have to, to not be afraid to change people to the right people. As quickly as possible, always without regret. Because again, it could literally last way past your lifetime.

Jon - 00:30:11: Yeah. And also, too, I think another thing that you said that was incredibly important is that you are doing yourself a disfavor and that other person a disfavor.

eMalick - 00:30:19: Yeah. Like it

Jon - 00:30:21: is not fair to them. Right. It's like because they don't deserve to have that, you know, tension. Like you like everyone wants to be in an organization where you find alignment. So just pretending like there is alignment is that's the disservice there. And it's actually you're doing the favor by being like, hey, it's not a good fit, but I will help you. I'll help you find a fit. It's just not here. And that's OK.

eMalick - 00:30:47: And there's another boat somewhere else. There's another boat.

Jon - 00:30:50: Tons of boats. There's tons of boats everywhere. And we will help you. It's not a fit here, but I will do my best to help you find that boat. And it's not a knock against you. It's just like, hey, look, we're doing something weird and different here. And we don't want any of that heartache. We don't want to discover this heartache later. Let's figure it out now. So, okay. I'm going to get off my soapbox for a moment. So now you're like, okay, I need to go in my direction, my own direction here. Can you talk a little bit about what was next after 

eMalick - 00:31:20: genetic

Jon - 00:31:21: intelligence?

eMalick - 00:31:22: Yeah, so I'm gonna... Shape this one sort of as terse language as possible. I had sort of two forwards like outward phrase in me and like a inner phrase in me. So leaving genetic intelligence, full of my shares, the most difficult sort of experience. I don't wish I'd find anybody, but I did it successfully. And I had more zeros in my bank account than I've ever had. Yeah, a little bit different now. I had more zeros and I was exhausted and I was sick of them. You know, with any people within the healthcare space. Just because of experience with genetic intelligence. So I needed like a break. I formed this company called Neurostorm, where we explored the frontier of of technology in medicine. Long story short, they put like a pretty amazing team. We worked together with Cornell University's Department of Surgery to explore different applications to help with Parkinson's disease, which ended up resulting in like a preclinical trial with people here in New York City that we published. And actually it was pretty successful in that sense. But Neurostorm itself never took off. And so, again, to anyone that's listening and trying to be an entrepreneur. Most things you do are not going to work out. And my example is Nearest Storm. But in the course of that, a friend of mine... Her name is Anna Rupert. She's a visionary MIT engineer. That I'm you know, was already like pretty successful as like a teacher. She went into, um, it's like a Dean of like physics, I believe, at one of the, um, academies here, um, came to me and it's like, some of the, some of the technologies that you, that you've been working

Jon - 00:33:10: with in,

eMalick - 00:33:11: Neurostorm particular virtual reality could be really helpful for this idea that I have in bringing, um, education onto three-dimensional space. And we could use some artificial intelligence to make this more, make more sense and be more, more useful to, to, um, for learning. Um, so this is completely outside of my space, but I was like, that's such a vision. Actually will help you out on this. Also, I was motivated by when I was in graduate school, we were studying brains, like literally physical brains that were cut into these different slices. They're like, so I just meat all over, like put on a table and you go study it. We could do better. 

Jon - 00:33:47: Yeah, yeah, yeah, yeah.

eMalick - 00:33:49: It smelled really bad and it just looked terrible. And I wanted to move us forward into the future. And this is my opportunity to have an impact on the educational system. So Anna Rupert would come to my office or would come right here to this living room, also her place, and shop the company and build it out. And eventually it was strong enough that she could go. She went out on her own. And now Prisms is. And, um, I think at least like 40 to 45 different states, hundreds of school districts across these states. So it's not Andreessen Horowitz funded. So that was like a stunning success. But it came out of the fail of Neurostorm.

Jon - 00:34:36: I was going to say for listeners out there too. These quote unquote failures, it's all a creative, it's all a creative, it's all a lesson, right? It's all a lesson. If you're willing to make it a lesson, right? I think a lot of the time, you know, you get whatever your, your Delta failure, your, it doesn't work out the way you want it. And. You know, you basically, of course, take time to mourn it. Like for sure, mourn is like a painful as hell experience, but there's always something that you can take away from it for your next, whatever your next thing might be. But if you let it just all consume you and wallow and mourn it forever and just not take away any of the lessons, that's when it's like becomes problematic. But if you don't, of course, take the time to mourn, but also see what is the kind of the silver lining to it. Because like, look at this, prisons of reality came out of this. Right. And that's like, right. But if you could have easily just let it, you know, just like pound, like pound you into the ground and you're just like, I just give up. So I think a lot of the time entrepreneurship, and again, this sounds like trite and super just cliche, but a lot of it is just like gritty, not giving up.

eMalick - 00:35:48: Yeah, in fact, it's resilience. It's resilience and... It shows that you're trying. That's how I look at it, right? That's my silver lining. It's like, if you don't have any fails, that means you really didn't try hard enough, right? Yep. So you're not exploring the boundary space.

Jon - 00:36:04: Exactly. And I think I was speaking to Nathan Clark. He's the founder of Ganymede. And he was talking about something about it. It's like, there is no... Kind of like silver bullet or just like. Insight here that's all too special. It's just like there are people who are doing things or not doing things. And. I want to be the person that's like doing things. And that ultimately is usually the indicator that things are going to work out. You just have to keep doing and

eMalick - 00:36:33: keep

Jon - 00:36:34: it pushing and just keep moving forward. So, you know, I love that prisons and reality came out of this. So what

eMalick - 00:36:41: was

Jon - 00:36:42: after prisons and reality?

eMalick - 00:36:44: Well, so actually during Neurostorm and Prism of Reality, so that's what I had this front face and, you know, this company's sort of going forward. But I also had all these zeros in my bank account. Yep. So now I could do independent research. Cool. Without needing a company or university or anything, because if I want something, I could just pay for it. Bye. You know, dirty GPU is spun up.

Jon - 00:37:10: You won't get flagged by the admin.

eMalick - 00:37:12: No one's going to flag me. No one's going to flag me this time, guys.

Jon - 00:37:15: Yeah, yeah, yeah. You won't get 

eMalick - 00:37:16: flagged by Columbia. I'm not at the state velocity. You can't pull me back, right?

Jon - 00:37:21: Yeah, yeah, yeah. 

eMalick - 00:37:23: So I actually spent most of my time during the neurostorm and prism period looking at AI and genetics in the background. And I call them the hidden period, hidden or latent period, where I just went at a deep examination. Of what it would take for us to be able to identify the genes that are causing different diseases. I was very heliocentric on Alzheimer's and so forth, but that period moved me towards... Rare genetic diseases in general. And was also informed, I had time to process what occurred at Columbia. Us giving these, these rescue experiments, injecting these animals with the correct versions of genes, and then having these animals sort of move from being from a sick state to a healed state. And we weren't just doing it with just some disease, we're doing it with all different types of ailments. And it's like, what is the parallel here? Rare genetic diseases, right? There's about 10,000 of these diseases that affect one in eight human beings. So like a billion people across the world. And so I went into this deep dive onto some of the experiments and how we discover those genes at Columbia to see how we can move that. In humans. And again and again, the answer is like AI. You need to use AI to be able to process this amount of data, right? But the AI itself, you just cannot make like a transformer and do it. It has to have some sort of framework that allows you to repeat what we were doing in C. Elegans, but in a much more complex, high-dimensional human genome space. So I got into different groups of studies that I wasn't exposed to before. So one that really sticks in my mind is coalescent theory. Coalescent theory is just looking at the movement of genes through populations. Like a good example would be... I use a blue eyed, green eyed gene. Denmark, Swedish, like blue, one of them is blue-eyed, the other one's green-eyed. And then you have two people that lived there a thousand years ago, and then one person got to migrate out of that space and move to Italy. But everybody in Italy has brown eyes. How long will it take for a Danish or a Scandinavian person to have a kid with an Italian person? How many kids do they need to have for them to have the blue-eyed, green-eyed gene show up in a population? Okay, that's like Mendelian genetics, right? But how can you describe that spread across a population over multiple generations? So, really esteemed researchers studying birds and looking at the movement of genes across bird populations over time. There's folks that are doing it with people and just looking at general features in people across time. But there's also with diseases, just looking at a particular genetic element. Have examined its movement across time. So the Alzheimer's gene, for instance, you could track the Alzheimer's gene from 1,100 years ago in a particular family and then look at the family records all the way until today and see those individuals are now distributed in five different continents, but all have the same change in the gene that causes their version of Alzheimer's, right? These techniques are wonderful, but they haven't taken over. Because they're very manual. They exist in a very low dimensional space. There's no doing any of this at scale. No doing at scale. This is why AI is needed for us to be able to discover genes that cause different elements. AI systems, by definition, look at multi-dimensional spaces in a non-linear way. Right. The human genome is multidimensional and it's nonlinear. We tend to think of genes as like, you know, five prime to three prime.

Jon - 00:41:22: Yeah. Yeah, yeah, in the textbooks. In the textbooks.

eMalick - 00:41:25: Yeah, in the textbooks. And like for my direct experience in the lab, you know, you have like this. Part of the genome called a gene and it's being impacted by things in front of it as well as on the other side of it. So it's quite multi-dimensional and AI systems excel at problems such as this. And during that hidden period, you know, where I'm just blasting away at my GPU as much as possible. It's such a wonderful freedom to have, you know. Yeah. I also started to see that... Something that, you know, it's started a dovetail into today that like AI systems, when designed the right way, could start looking at the human genome as a first language. What do I mean by that? So the first language is like, you know, we speak English right now. And the second language or third language could be like, I hear Spanish all the time, but I really, I could say maybe a few words in it, but I can't, I'm not really fluent in it, but I understand that the language is being spoken and I could dice apart. Well, if somebody's cursing at me, I get that pretty quickly. Yeah, yeah. But like a first language is something you sort of intuitively, you don't have to think through the word, you just get it. Today we operate... With the human genome, we cognitively see it. Or absorb it as like a third language. The experiments we do are all third language experiments. For instance, to discover genes in the laboratory, my group, as well as all others in the world, would design experiments where we essentially randomly mutate the genome at different positions and see the change in the animal's ailments and then assign that gene to that ailment, you know, for genotype, phenotype. So we moved from that to doing targeted sequencing with CRISPR, but it's the same idea. It's just one by one movement across the genome to try to understand meaning and just assign words to the genome. We don't look at the ATGC code and be like, we get that, that just means that. We just don't, right? We can't. We don't believe it.

eMalick - 00:43:43: We haven't evolved to that level of intuition.

Jon - 00:43:46: Yeah.

eMalick - 00:43:48: So what I saw during that hidden period was some hints, some hints that is a potential for that to occur. Using AI systems. So one model that I had, also I have to say, all these models, creating them from the ground up, right, which is, again, like, but it was something I was doing at Columbia when I was like a, you know, a bit of junior learning how to program. So now, so ground up, none of this like GPT wrappers. Yeah. You know, wonderful companies are going to come out of that, but we design things in the ground up because we have to. We work on the genetic code and things have to be custom for that. But yeah, we designed this model that, you know, that was able to group different heritages from all around the world without it being given any labels of what those heritages are. So it understood the difference between Swedish people and Australians, Japanese people and Madagascans, Austrian people and Gambian people, where I'm from, folks from Texas versus folks from Argentina. Without giving any labels, just a pure genetic code. Man. In fact, one of the surprise findings of that study was just, again, the pure genetic code, we're able to find a microheritage in Japan. So I had thought, we were talking about Japan earlier, I thought the Japanese people were just this uniform. It's like a Japanese group, just basically one people, so Japanese people. But it turns out that Japan we have today is a result of multiple different migrations of people from mainland China. And the earliest of these is a group called the Ainu people, A-I-N-U. And within the AI model that I designed, there's this massive cluster of... We use cluster so humans could understand what the AI is thinking. 

Jon - 00:45:42: Yeah. The AI... 

eMalick - 00:45:43: Yeah, there's this massive cluster of Japanese individuals, and there's this micro-cluster right next to it. And it turns out the micro-cluster is associated with South Korea, with mainland China, and some of the Pacific, North Pacific islands and so forth. But the model separated these individuals. It's like these are genetically... Japanese, but genetically different from most Japanese. Interesting I was like, okay. Yeah, when I dug deeper, it turns out this is Ani people. They were the first wave. They're sort of the original settlers of Japan, in like North Japan, Japan, like Aiko. And there's only about 200,000 of them left now. They have their own language, which only like less than 100 people speak today. So it's like an extinct or near extinct language. So all of this cultural information came out of the finance of the model. The model itself.

Jon - 00:46:43: That's trippy. That's, and so as you're like, you know, you're doing this research and you've created this environment of like. Creative freedom, you know, you're kind of truly like liberated in that sense. As you're looking forward for it, I know you've been getting active. When did it become from a kind of independent research to I'm going to like turn this into a company?

‍

eMalick - 00:47:08: Yeah, so I was waiting for a few signals. I knew eventually the next company is going to come. I need to, one, sort of recover from the genetic intelligence, not understate that I'm a person just as much as anybody else. Two. Pandemic occurred and we wanted to wanted to get out of that phase of fundraising to sort of come back online and three there were some limitations within the convolutional nets that like some architectural change needed to occur for them to be able to take even more higher dimensional genomic data. So I recovered mentally and physically, the pandemic changed and the transformers came online. So I'm not sure how you're familiar with transformers, but they're essentially what's. 

Jon - 00:47:59: For everyone out there who might not be, and I only know it at a very surface level, so I could benefit from learning from you here.

eMalick - 00:48:06: Yeah, absolutely. So the convolutional neural nets were what drove the AI space from 2012 to, I would say, 2020, right around the pandemic. And they have this very specific architecture that was first fine-tuned on vision tasks, but it's actually broadly applicable. However, it has a ceiling as to how much information you could put on it. And people certainly tested the heck out of the ceilings. A new type of AI came about, I think, 2016 called a generative adversarial network. Which... Again, took level up the playing field a bit, but it also had issues that needed to be overcome. In 2017, a paper came out from, I think it was actually a 2016 paper, came out describing this architecture called a transformer. And then folks at Google, now part of those folks that are co-founders of OpenAI, describe that model in much greater detail in a paper called Attention is All You Need. And they described, basically used this type of data representation called Attention Framework in an architecture called a transformer. And fundamentally, the issue that they were looking at was how do you have any, like within like a chain of words or a chain of tokens, whichever way you look at it, How do you at any one moment within that chain? Be able to look back at only what's important behind you. So convolutional nets just look at everything and sort of give it equal weight, while within a transformer network, it's able to sort of be more judicious about what it looks behind, sort of in a look-back table, so to speak, right? This is like a very low-key definition of...

Jon - 00:50:02: Yeah, no, no, no. No, that's exactly how I need it, because it sounds like that's a massive efficiency gain. 

eMalick - 00:50:08: Yeah, and this was a game changer. And, you know, these things take some time to matriculate into different uses. But around 2020, 2021, these transformers started being used for all sorts of different tasks, from developing GPT-2 to GPT-3, also being adapted for vision tasks. They essentially took the cake of the generative adversarial networks. They just did what they were doing with much greater representation and much more robustness. And I am always looking for architectures that could take in more information because the goal that we have is to put in thousands of genomes within some architecture. So remember, first we did that with a single genome in a convolutional neural net. And now I want to do that to thousands of genomes so that we could get this understanding of the genomic space as like a first language. So when those three signals occurred, I recovered pandemics over, transformers came online. T-Cartoon was ready to be born.

Jon - 00:51:13: Star of the lining. I can see it.

Jon - 00:51:16: Very interesting. So now the timing is like, just absolutely just like, you know, it seems like this is like, it's starting to just like, everything's like aligned.

Outro - 00:51:26: That's all for this episode of the Biotech Startups Podcast. We hope you enjoyed our discussion with Emalick Njie. Tune in to part four of our conversation to learn more about his journey. If you enjoyed this episode, please subscribe, leave us a review and share it with your friends. Thanks for listening, and we look forward to having you join us again on the Biotech Startups podcast for part four of Emalick's story. The Biotech Startups Podcast is produced by Excedr. Don't want to miss an episode? Search for the Biotech Startups Podcast wherever you get your podcasts and click subscribe. Excedr provides research labs with equipment leases on founder-friendly terms to support paths to exceptional outcomes. To learn more, visit our website, www.excedr.com. On behalf of the team here at Excedr, thanks for listening. The Biotech Startups podcast provides general insights into the life science sector through the experiences of its guests. The use of information on this podcast or materials linked from the podcast is at the user's own risk. The views expressed by the participants are their own and are not the views of Excedr or sponsors. No reference to any product, service or company in the podcast is an endorsement by Excedr or its guests.