The Business of Connective Tissue w/ Shahar Keinan | Ep. 85

In this podcast, David and Gary speak with CEO and Co Founder of POLARISqb, Shahar Keinan. The conversation takes us to CHATGPT, medical innovation, quantum computing and some of the best advice for transitioning from a technician to a leader that we have featured on the podcast to date.

Links:

https://scholar.google.com/citations?user=vsQGltwAAAAJ

Shahar Keinan LinkedIn

POLARISqb Website

POLARISqb LinkedIn

___________________________________

Submit Your Questions to:

hello@thebigpixel.net

OR comment on our YouTube videos! - Big Pixel, LLC - YouTube

Our Hosts

David Baxter - CEO of Big Pixel

Gary Voigt - Creative Director at Big Pixel

The Podcast

David Baxter has been designing, building, and advising startups and businesses for over ten years. His passion, knowledge, and brutal honesty have helped dozens of companies get their start.

In Biz/Dev, David and award-winning Creative Director Gary Voigt talk about current events and how they affect the world of startups, entrepreneurship, software development, and culture.

Contact Us

hello@thebigpixel.net

919-275-0646

www.thebigpixel.net

FB | IG | LI | TW | TT : @bigpixelNC

Big Pixel

1772 Heritage Center Dr

Suite 201

Wake Forest, NC 27587

Music by: BLXRR

Transcript

David: 0:05

Hi, everyone. Welcome to the biz dev Podcast, the podcast about developing your business. I'm David back to your host, and I am joined by professional Spanx model. Gary Boyd. How's it going, man?

Gary: 0:16

It's going pretty good. I actually just got my contract renewed, but I'm under an NDA. So I can't really turn around karega

David: 0:23

Can you breathe okay with all that holding in to what you don't know is Gary's three times that size.

Gary: 0:29

It's not my breath it it constricts the but yeah, I gotta cut down on carbs, man.

David: 0:35

He is three times as Spanx are really good. They just really suck it on in there. Anyway. More importantly, we are joined by Shahar Keenan, who is officially the CEO and co founder of Polaris qB. So we're gonna learn what all that fanciness means here in a minute. I wanted to talk about briefly I want to start our conversation. First off, welcome. I was gonna I don't do that enough. I don't welcome our guests enough. And I'm just rude and just dive on in there. So welcome. Okay, so I've done that. She smiled at me. So I'm gonna count that as a win. I wanted to talk about this little thing I found on LinkedIn, it's about AI stuff, because we've been talking about this, and this is what's hot. What did it say? It said, I want to get this right. I sent it to my team this morning. It said that chat GPT for the newest version of open eyes, chat GPT AI bought thingamabob er that we've been talking about just past the US medical license exam with flying colors. More importantly, it diagnose a one in 100,000 condition in seconds. This guy says the future of medicine has changed forever. That was a LinkedIn post. So you're in this space, right? This is where you live in the medical kind of spacey space. What does that what does it? What does that do for you? Does that scare you? Does that excite you? Neither? What do you think?

Shahar: 1:56

It excites me? Technology excites me to begin with? But also, come on, we've been through enough of those hype cycles, right? So it's exciting. But unless we really understand what it does what it doesn't do, we have to be careful there. Right? We're not we should not expect people to come to their physician tomorrow and say, you know, I have a stage four terminal cancer, how come check GPT doesn't find something for me. Right? It is more and is less than what we are saying it can do. Can, what I where I work for is designing new molecules, these molecules can become drugs, the process of making a molecule into a drug is a seven year. Right. That's that's why we have all of these big pharma companies out there. And they take the output of what I do or other people in biotech do. And they take those molecules and make sure that they are doing what they're supposed to do so cure certain disease, and don't cause too much damage on along the way. And, and that's something that at some point, you can only do that by trying it on real people suffering people. And you have to make sure that the drug that you give them, cure them or lower their their effects and don't cause too many side effects. And we have regulatory FDA that makes sure about that. Right? And we have a whole sort of system to support that. Now, churchy, PT Can you can ask it, should I call it? You can. You can ask it a thing like you ask it a question. And if somebody else knows the answer to that question, and put it online, it will give you the answer. Yeah, if you ask it something that nobody has ever asked before, or there's no information online, it will give you garbage. How do you know if it gave you garbage or not garbage? You usually will go to somebody who knows the answer and asked, right. What would have happened if we would have asked Chuck GPT something completely spurious. I would have asked it about some disease that I just invented. It would also give me an answer by the way, I've tried to

David: 4:39

give you a very confident answer. Absolute gibberish. Yes, yes.

Shahar: 4:42

So so it works. In some cases. It works beautifully, amazingly. But it also you have to know what to ask how to ask. I don't think it's going to change it. It is changing parts of the world. Have of where we live. But it's gonna take a while.

David: 5:03

But I can totally imagine. And they actually had a television show, I'm addicted to New Amsterdam currently. And they had an episode couple years ago, where they had a tablet that one of the doctors was carrying around, and they would hear the symptoms and complaints from the patient. And they would turn around and help the doctor diagnose. And I can totally see that changing medicine can vary, because then you got the world's most knowledgeable doctor, right? They've read literally every medical book ever written. And that's at the fingertips of every doctor, some of which are horrible, some of which are amazing, right? So now you've got this diagnose diagnostician boy, I just butchered that. I just killed that what a person who diagnosis it right, that's a whole that's a whole type of medicine, right that diagnostician or whatever, doctor, it was really make their lives a lot better. Right? You You have Dr. House who was crazy and funny, and TV, but those people exist, but now you don't need a weirdo like him on the TV show. You can have all of that knowledge in a thing. Does that change medicine? I think it does in a way.

Shahar: 6:16

But you know, medicine is statistics. Right? Right. You know, when you hear hoofbeats? assume it's horses, not zebras, right? That's medicine. But there are the zebras. Right. So when you come to your physician, and you're coughing, she will say allergies, called flu, and she will go down the list, right until at some point, there's some other things there. And, and you are, you know, and that's what chatty PT does, as well. And maybe GPT is really good at saying this is allergies, because this is the right season. Maybe that thing doesn't know how to go through the lower cases, the less statistically relevant cases, because they are not so much documentation of them.

Gary: 7:10

I do remember reading headlines and seeing articles about how AI is assisting with creating new sequences and proteins to help with coming up with them, you know, either vaccines or other molecules. Is that something that you see working well in your industry? Or is it just a lot faster than humans in that area where it's just giving you a lot more to choose from and test?

Shahar: 7:35

I think it's the the second option, okay, it's good. It's faster than human and it giving you more things to test. The issue is, think about the difference between incharge GPT, between version three to version four. Right? So so they added, I thought it was like two orders of magnitude more data in order to build the model. Right? It's an all it looks, maybe it looks very smart. But it is by the end of it's a machine learning. It's a model that you built, you take the data, you digest it, and then you ask questions and the questions. The answers have to be something from the data. In, in, in my areas in my field, drug design, biotech, protein design, there is a problem of not enough data. Okay. The deficits that we have are very small 1000s of molecules, which is a lot of effort in put into that, but that's not a lot of data. There are places where there's much more data, okay. For example, something that called what protein folding how, what's the structure of the protein? A protein is this is a long molecule made out of multiple amino acids. And it folds on itself, it creates a three dimensional structure. What's that three dimensional structure. So only in the last year or so there were models that are good enough. One of them came from me from through Google. Folding, there's one through meta, there's two groups. In China, there's another group in University of Washington. And all of these groups basically have enough data, experimental data, that somebody built the model, sorry, built the proteins, synthesized it, measure it, and put this into a database. And then these groups came in and looked at that database and created the model. And only in the last year, these models are accurate enough that we can start using them. I want to

David: 9:47

back up like I always do. I feel like that's my new soul. Right Gary, now it's backing up anyway. Tell me about your company because this is ultimately a business podcast and I want to make sure that we Talk about your company and how you how it came to be your co founder. So where did this idea come from? How did how did you guys come to be?

Shahar: 10:08

So I'm a computational chemist by training. So a scientist, worked in molecular drug design, for many years, looked, one of my things that interest me was something we think of as chemical space, what's the possibilities of connecting atoms to create molecules, and make something useful out of them? Okay, like drugs. About three and a half years ago, Bill Shipman, my co founder, and myself, we were working at a different company, that company did not succeed, or succeeded for a while, and then was less and less active. And we were sitting, and we were trying to figure out, you know, what do we want you to do when we grow up? What kind of what did what didn't work before? And can we fix it. And one of the issues that we found was that the number of molecules that we can look, it was not very large, even on a really good computer, you know, Google Cloud, we can have access to 1000s 1010 1000s of nodes. We wanted to have a bigger chemical space we wanted because, okay, let's step back, one wants to drugs, molecules, they need to do two things, right, they need to cure you, and they need to not kill you, at the same time. Okay, that's, that's what you want from a drug, the cure part is usually bind to a clotting machine in the body. And either make it work smart or worthless. That's, that's how most of the drug works. They bind to a machinery in your body or in a virus body, and either stop it, or make it work better. If it's a, it's in cancer, you want to stop replication of the cancer cells. Right? That's so you find a protein in the body that replicates the cell, and you stop that replication, okay, maybe you have a heart problem, some some machinery doesn't bring something into your cell, you want that machinery that probing to work more to bring that nutrient to the cell more, okay, so these are the two kinds of things. Usually, vaccines work in a different way, these are not vaccines, these are drugs. Okay, and so, and then you so you've wanted to find, so when we do drug design, we find molecules that bind to the protein, and either makes it worse, more or worthless. But we also want that small molecule to have other properties. And those other properties are everything from how easy it is to make it to not interact with other things in your body, and do something bad. And you can think of this basically, as a multi object optimization. So you're looking for a molecule that will have some properties. So there are some objectives you want, and some constraints. And the way to find these molecules is to look through chemistry, what are the molecules, and it's not easy to find those molecules. Okay, it's easy to find molecules that either don't bind to the protein, or if they bind to the protein, they bunch to a lot of other proteins. So finding those molecules that bind to only the protein that you want, and don't have a lot of side effects. If you think of the Venn diagram of that, that overlap between the two circles, is quite small. And you need to look at very large circles to find. So we were sitting there and saying, how do we find those molecules? What technology can help us to find those molecules, and the technology that we are using is using quantum computers? Specifically, we're using a quantum annealer, which is a kind of a quantum computer that is very good at optimization problems. We use machine learning but a bit differently. Yes.

David: 14:41

I forgive my ignorance against small brain. I thought the quantum computers were mostly not real. Like they weren't there yet. I mean, I know they're like the next stage right? Where we there's all sorts of craziness that those bring to the table because there's they're fundamentally changed the way we do computers, but But I thought they were mostly science fiction at this point, you're saying you guys use them and this is a thing.

Shahar: 15:05

So they're not. They're not science fiction. They, they're, most of the quantum of the quantum computers today work, but they're very small, they have less than 100 cubits. Think about what kinds of codes you can write with 100 bits. That's what you can do there, you can do some simple problems only. But there is one kind of quantum computers, and can solve only a single kind of problems. Okay, so and that's quantum kneelers. Company, like D wave has a quantum annealer. And it can only solve optimization problem. It's very good at solving the backpack problem, they are traveling salesman. It's not good at anything else, it cannot do anything else. It cannot do hello world, it cannot do factoring of numbers. It can only solve one kind of problem. Think about GPUs, even more limited than GPUs. What kind of problems. So if you have a problem that you can rewrite in a certain format, it can solve it for you faster than any other system. But only if you can, if you can take your problem and reformulated there. That's what we do. We reformulate chemistry into the language of quantum computers into cubits specifically, which is the algorithm that we're using.

David: 16:50

So in the ones, the ones that I've read about, the idea is a regular computer does zeros and ones. So it's binary, right? That's what all of our computers run on, right now on and off. And there are many theories of how quantum computers work, which because from what I understand the science fiction version is, there are 32 states of an electron, therefore, you could have a set of zeros and ones, you could have zero to 32 different versions of it, which would be stupidly powerful. The ones that I've heard of in the read about that are real, and I'm using air quotes there, because in my mind there, they don't do anything, but I'm wrong. Obviously. They do three, right, is that is they have three states. That's the qubit, isn't it?

Shahar: 17:35

So so they do. Okay, again, the quantum computers like the Gates computers that IBM or Google or, or iron Q or atom, they can do stuff, but it's really small systems. So so if you are thinking, What can I do with those today? The answer is not a lot. What can you do with them in five years? If the engineers will solve all their problems, right? Of course, we always blame the engineer.

David: 18:09

Sure. Well, that's always what do I blame the designers?

Gary: 18:13

I blame the devs.

Shahar: 18:14

Yep, yep, each one blames a different group. But but if they can do that, then we will be able to solve many problems that today takes a really long time they take so long that you cannot solve because of that. Quantum computers, there is, again, the D wave that kneelers can solve one kind of problem. And they are very good at.

David: 18:40

So how did you get to the point where I have lots of questions here. Yeah, not to nerd out too hard on the quantum stuff. That's just fascinating. But how did you get to the point where I'm good at chemistry, I want to do things. Let's do quantum computers. Like that seems like a big jump where did

Shahar: 18:56

so we did a whole lot of searching a whole lot of thinking. It we had other options as well. And we were also a bit lucky. We got introduced to a company called Fujitsu, who have a kind of a quantum computer, and they were looking for test cases. So our first proof of concept was sitting and talking with them. We've tried other things. They did not work very well. We did try the machine learning route, which we found, especially here for chemistry. I think there is not enough information out here. There might be companies that have those information, enough of it. But public knowledge there is not enough. What what is what you can get from public databases. It's definitely not enough.

David: 19:49

So who is your customer?

Shahar: 19:54

I have two kinds of customers. I have customers that come to me and say We have a difficult problem, design a drug for us. So these are partners. They tell us everything. You know, what their protein, what's their disease, what the constraints, what their objective. These are usually small and midsize biotechs that have a problem that they cannot or that it's difficult for them to solve. And they come for us for that. That's one kind of customers, we are also developing now a SaaS product. And that's going to be a very different set of customers. If the Baltic customers have one to three projects per year, the SAS products make much more sense for bigger companies farmer that have maybe 10 projects per year. But they don't want us to touch their data, they don't want us to know what their targets are, what's their molecule, the outcome of the molecule, or any of those things. So this is much more as a SaaS product, okay, they come in to a website, they submit their queries, it goes in the background, they come back tomorrow morning, and they get the results, the accuracy of their results is not as good. They get maybe a 10,000 molecules. When we work with biotechs. With small companies, they get 20 to 50 molecules, it's much closer to what they need to do the next steps. Pharma companies get much bigger chemical space. But they can take it in there internally and find the best ones in there. So these are two different levels of complexity of answers to different level of engagement with us specifically.

David: 21:44

Wow. Okay. So what made you guys choose? I mean, you're making molecules, which means you could, in theory, make anything right? That's where we're all made up as molecules, right? So what made you go drugs, as opposed to plastics, or energy or anything else materials.

Shahar: 22:03

So this is where our own expertise came in. We are drug designers, we worked with biotechs, before we worked, we knew the market when you the consumer, when you come to me and say what is a drug, I know what's a drug, I know how to design drugs. I know how to be a seller in a drug market. So So that's they, they there's a technology, right? But there's the the our knowledge, our experts knowledge, and that's, that's, we took a problem. And we find a solution to it. Using a certain technology.

Gary: 22:45

This seems like a very specific business model for a very specific group of clientele. So I'm guessing you already kind of had a client or targets for clients when you were coming up with this idea. And then did you reach out to them? Or did they know what you were doing? And say, I can't wait to you guys get set up? And we're gonna come to you with our questions, or I mean, how do you how do you market or try to reach more available clients with something this specific? So it's, it's not

Shahar: 23:13

as unique? So yeah, there's not millions of customers out there. But there are a couple of 1000s. Maybe, if you say, I'm not just looking in the US, I'm looking all over the world, because biotechs you can find all over the world, there's, that's the market. They there are marketplaces. So and these are usually more of a partnering conferences. So there is a partnering conference, everybody goes to it. And you say this is what I'm selling. This is the kind of, of customers I'm looking for. But the customers come in and say, This is my problem right now. And they can come and say, I'm looking for somebody to design my molecules. And they can also come in and say, I'm looking for somebody who can test this, or somebody who can give me this kind of service. Okay, so there are many services. You know, when we design, it's not enough to design the drugs, you have to test them before you can go to the FDA and say, I want now to test these molecules in people. You have to do many tests before the so there's a very large marketplace of things that can happen before a molecule become a drug. And before the FDA will let you try it on people.

Gary: 24:29

So it seems like this is already an existing network that you just once you get in, you're in and you're kind of connected with everybody in the same space already.

Shahar: 24:38

Yes. So there's several marketplaces like that or several conferences that you can go to. And they are they are how you market to in those places. It's not you know, you always we always learn I started as a as a scientist and if Anything I've learned in the last three years is how to be the the business side of things, right? How to be the CEO. So how to find customers? How to find the investors, how to find team members, everything else, right? How to how to build a company, how to make it a company that will continue to be successful? Yeah, and customers is definitely one of the most important things, will you

David: 25:32

you're leading right into a question I have. And that's where you're in a classic situation where you have two technicians starting a business. Neither of you are business people, you got the two nerds, right. And how, and you already touched on this, but how do you go from two nerds to a really good at what you do clearly? to running a business? Like that's a very different skill set.

Gary: 26:01

And we mean, nerds in a very loving way.

Shahar: 26:04

Oh, yeah, yeah, I'm a proud nerd here completely. And

Gary: 26:09

listening to everybody,

Shahar: 26:11

everybody in Polaris is a nerd, because, you know, we geeked out on technology, we geeked out on solving technological problems, biological problems. So how do you do that? I think people who are nerds love to learn, okay, we just we're learning we are good at learning. We that's what we do. Right? That's part of being a nerd is is your you just you learning. And this is just learning other things. And we learned a lot, we have really good advisors. We are in a good area. Okay, we are in the Research Triangle Park in North Carolina. So there's a lot of people around us, who are the combination of nerds and also involve on the business side of things. We are in a in an in good environment. So the American underground. And we just sit down and learned we've read a lot. We talked with people all the time and continue to do that. We listened to our customers, we listened to the people who said no. Many people say no all the time. You have to listen to them, as well, as much as you have to listen to the people who say yes. Right? Why would somebody say no to you? I mean, I bring them the most technologically advanced solution that they will ever see. Why would they say no? And when you talk with them, you understand how to make your process better. Your sales process better? Your product. Okay, so you learn both from the yes and no, your must learn more than from the No.

David: 28:11

It's just it's really interesting. I mean, this is like a mini masterclass on how do you go from technician to running the business because a lot of people who are just pure Tech's would regardless of what they're they're into, they get paralyzed by the social slash businesses side, because they're like, Well, I don't like people. Alright, that's common and nerd circles. I don't want to talk to people, I need to find and partner some with someone who will be that front. But it seems like you guys just said, well, everything's a problem. So I'm going to solve it, like I do every other problem, and just learn and learn through it and get over my own predilections for being quiet or not being around people or whatever. And that's just a really fascinating, unusual way of hearing to most people, once there, there's a problem, they're either going to find someone to do it for them, or they're going to realize they can't do it and find a workaround. And you guys, like now we're just gonna learn how to do it. That's really interesting.

Shahar: 29:15

But also there is another thing here, right? If you go and maybe maybe 30 years ago, okay. There was only one way of doing things right. When you sold something. It was almost a you know, used salesman kind of things, right? Today, you can find ways to do things that are more in tune with who you are. Right? So so when I go and I talk about what we do in Polaris, I talk about technology. I connect with people through their problems through how to solve their problems through our technology. It's you You need to find how you how you solve your debt problem, are you so when I say, you know, I wanted to learn how to be a best a better seller of what we do? The answer to that was to find other people who are selling for technology that are like me, right? I cannot go into a room, stand on the table and say, Listen up, guys, that's that's not who I am. But when I sit and talk with people, I can, I am much better at that. So you need to find how to do the things that you are not comfortable with. Using your own, you know, your own, what you are, right, how you how you sell how you manage how you talk about the things you do, using your own strength. Each one of us can talk about the things that you love, right? Even the most introvert, you find what she likes to talk about. And she will talk with you and you can connect, and you can maybe solve a problem together, right? Because by the end of the day you solve a problem, nobody's going to buy anything from me if I'm not solving their problem. So you just need to find how you do it, how you find how you yourself can solve it.

Gary: 31:29

That's an interesting take, you know, a lot of times people will say you have to learn your limitations, and then get someone else to do it for you. And it sounds like what you're saying is when you reach those limitations, instead of trying to adapt to the situation, you kind of let the situation adapt to you and meet you on your own terms, which and this kind of leads us into our question that we asked everybody that joins us on the show. What are your top three pieces of advice for any entrepreneur or new business when they're starting up?

Shahar: 32:02

So So three things. The first one is when you have a problem, right? You can look vertically, you can look at your competitors, your customers, other people in your area. But you should always also look horizontal, what other people are doing that got nothing to do with you. But maybe solve your problem. Right? If you know Chuck GPT didn't start a church up, they probably took something from somebody who did something at Vision, right, or different kinds of machine learning. So you you don't look just horizontal, sorry, don't like just vertical look horizontal. We are of course we are all individuals, of course. But at the end of the day, in many cases, somebody have had your problem and solved had a different problem. And you can use that technology to solve your problem. Always look outside. That's one. The second one is the answer is always personal. How you sell how you buy, how you develop something. These are people and you need to always find the connect with the person that you're talking with. And then you can solve a problem, sell something, buy something, it's always personal. And the third thing is going back to the what we just talked about, right? You need to identify your strengths and your weaknesses. And either change the situation for you so that your strengths or your weaknesses become your strengths. Or you find people that compliment you. Okay, I can write some code. I'm not great at it. I probably am not a mazing at it. I have people that are amazing at writing code. Right? So you need to find what you do. Find what's your strength, find what your weaknesses are. Find people who compliment you and that you enjoy working with them. Because again, at the end of the day, everything is personal. You will work with these people you need to like them. You need to enjoy working.

Gary: 34:39

I like that everything is personal take because even as much as people will say well, it's business not personal. At the end of the day, a decision was made. And that decision came from a personal place. So yeah,

Shahar: 34:52

personal.

David: 34:53

Man this is this is fascinating stuff. I apologize to everyone who's not a nerd This is probably like we went way too far in the weeds, but I just can't help but that's just good stuff. But I want to thank you for joining us so very, very much. This is a lot of fun. If anyone wants to learn more about you, or Polaris, how would they do that?

Shahar: 35:14

So they will go to our website, www dot Polaris qb.com They will connect with me through LinkedIn. So I'm sure Harkonen, you can connect with me in our through the Polaris website, or directly from okay. And I usually try to answer emails I tried.

Gary: 35:38

You seem like you're probably way too busy to answer emails, but

Shahar: 35:41

no, no, no, no. It's amazing what? How, you know? Yeah, personal, like, what kinds of interesting people you meet and you talk with. So yeah, please,

Gary: 35:53

we will include those links in our show notes. And there'll be underneath this video on YouTube too. So if anybody does want to reach out, you could just look below the video, or check out the description in any of the podcast channels, and you'll find those links as well.

Shahar: 36:08

And thank you very much guys for inviting me. It was a pleasure talking with you. And noting together.

Gary: 36:15

Yeah, very interesting, very interesting part of this thing you do with quantum computing. I'm gonna have to look into that more. I know David's gonna start researching that a little bit more now. That's what

David: 36:26

we need to start programming for them. Let's do websites on a quantum computer. Let's do it. I don't think that's gonna work. That's okay.

Shahar: 36:34

We just published a white paper on a really I'm not gonna say simple problem, but they is simplified problem of how you choose a menu item, right? Chicken and Waffles. You know, these kinds of restaurants have to choose one chicken from many kinds of chicken one waffle for many kinds of Whopper. So this is an optimization problem.

David: 36:55

That is crazy. Gary, if people want to get in touch with us, how would they do that?

Gary: 37:01

They can email us as well hello at the big pixel dotnet. Or they can reach out to us on any one of our social media channels were on all of them. And those links are also below this video and in the show notes. Well, thank you

David: 37:12

all for joining us once again. And we look forward to next time. Thank you again.

Shahar: 37:18

Happy to be here.

Gary: 37:19

Thank you, Shari. It was a great conversation. We'll see you guys next week.