Katerina Axelsson - Founder and CEO at Tastry

Welcome to another Founder Wisdom podcast. We have Katarina Axelsson with us today. She's a founder and CEO at Tastree, uh, does back and forth in between New York and California. She's building a company, uh, yeah, tastree.com. You have the dot-com, that's a good one. And it's quite special what the company does, the first principles, uh, model for human taste, smell, and texture preferences, which is an industry in itself, right? So, Kat, welcome to the podcast. Tell us a bit more about, uh, Tastree.

Thanks, thanks for having me. So our tagline at Tastree is, is we taught an AI how to taste. And just to make a distinction, because I know there are a lot of people saying, you know, they have AI companies nowadays, um, we're first principles model, um, we're not a foundation model. But we operate off two datasets we had to generate in-house. One is on the chemistry of products, the other is on the preferences of consumers. I could explain more about that, but you can imagine similar to a drug discovery company, what we do is we run simulations of chemistries, whether or not they exist yet, against consumers to see what the outcome would be.

Right, right. Yeah. Before we get into this, tell us a bit more about yourself because you've run that company for what, 8 years? Did you— go straight out the gates as a founder? Like, what's your background?

No, uh, we had a very long childhood. I mean, it was almost 5 years of R&D, um, to develop the technology before we even commercialized a product. So, um, I mean, it was not the intention. I was pursuing my master's in bioinformatics, and, um, I was acting like a mad scientist at my job where I was working as a chemist, and Everyone said I should have been a doctor, but I say, you know, if I were a doctor, I probably would have created Frankenstein by now, and that would have been a bad idea. Um, but long story short, Tastree came about because I had inadvertently invented an analytical chemistry methodology that was very unique and powerful, um, and that became like the foundation of the proprietary dataset we use to build the company around.

You're—

so now, now I'm a founder and I don't, I just, I joke, I just wear a lab coat for photo shoots. Um, we have, you know, like PhDs and chemists, um, who have taken over since then, but that's how it started.

Right, right. Um, and yeah, you've been like, I guess your first beachhead is wine, right? Uh, why was that? Have you always been into wine yourself, or tell us a bit more about that?

No, not really. I'm not really a drinker. I prefer stimulants like coffee. Um, wine is because, well, I went to Cal Poly and there's 400 wineries in a 40-mile radius, and that's where the initial support, um, came about. That's where the idea was born. Um, so it's just the first vertical we tackled, which was good and bad because from a chemistry and complexity perspective, I can't think of a more difficult vertical to prove the technology in. It's very complicated. Um, the market is unique and fragmented though, so we found out later that we picked the hardest vertical to, to build the technology and the hardest business to sell it in So everything after that became much easier.

Yeah, that's quite intriguing. And why did you pick taste specifically? Like, how did you get into tasting?

Uh, tasting is, so it's, I would say it's more about preference, I would say. So I'm odd in that I obsessively read research papers and because of the nature of my job at the time, I was obsessively reading research papers about all these multi-billion-dollar companies for the past 50 years who have been trying to map words to molecules to predict what consumers are going to like. So like, I kid you not, there's billions of dollars spent a year trying to predict, like, are you and I going to describe my cup of coffee as tasting like cherry?

Yeah, no, it's impossible that we both agree on the same thing, right?

It's like Right, and it's always a hit or miss because we have different cultural backgrounds, different experiences, taste buds, taste buds, food we ate previously, environment, smells.

Smells impact taste as well.

So, right, and so words are not really adequate for this sensory element in humans. And then I thought, well, wait a second, it's wasteful because even if you and I could agree— what, we have 100% accuracy and you and I agree this coffee tastes like cherry— you and I could still disagree on if it's good. Like, I could love it and you could hate it. Yeah. So I thought the, the actual problem is preference and how do you predict what people are going to like whether or not they can describe it. And that is predictive, and that's where it got really interesting. Um, I think I answered your question, but let me know if I didn't.

Yeah, but yeah, what's, um, yeah, my brain was like, why, why do you prefer coffee to wine, and what does that tell about you? Like, let's say I, uh, I prefer like strong black coffee as well, right? Like, especially when I work, it's going to be stimulants. Love a good wine every now and then when, when it's in very good company, right? Um, Do you also evaluate the impact of what it tells about someone to, I don't know, like black, hard, and sometimes, yeah, very sour and disgusting coffee?

Um, it's not always a one-to-one relationship, right? I know people who drink black coffee who love desserts, right? And they might like bitterness here and not there. And I think that's why it's a perfect problem for AI to solve, is because it's way more complicated than, uh, linear understanding of all these hundreds of interdependencies that are making your palate unique, right?

Yeah.

If it were so simple, we wouldn't need AI to solve it, is kind of the way I see it. We're so unique just in wine. I know there's 60,000 unique palates, 60,000 archetypes of people and the way they drink wine. Um, so yeah, scary, but you got to simplify it for consumers, right?

And I'm looking at, um, your website. So cherry, marzipan, nutty— like, what does marzipan need, right? Like, so the— because the consumer needs to have tasted various things to associate what it tastes like in the first place, correct?

Well, we believe consumers and people in general are very unreliable and tell us what they want. We don't— that's how it's always been done. And I'm taking more— where we are as a company, taking more of like the Steve Jobs approach, you can say, where it's like we're intuiting what people want before they know they want it. So not having much data on consumers at all, that data can be served as an analog to understand what you want better than you can is the way I would say it, right?

And your company, you take, uh, the data from customers, you take general trends in the market, and you feed it to your AI. Your AI returns, this is what their customers may want right now, and this is how you should describe it and brand it and sell it.

Or yes, that— yeah, basically any part of the supply chain that wants to predict and anticipate consumer consumer demand, but also we can turn an opportunity in the market that wasn't seen before directly into a recipe. And that's really, I think, the most pragmatic and profitable business application right now is— I'm so sorry, um, that it is, uh, how do I compress 2 years of R&D, millions of dollars, and trial and error where every 3 weeks I'm going to a focus group to see how consumers react to this, going and making another iteration. And by the time I'm done, it's been 18, 24 months and the product's in market. Just future-proof the outcome and get a better outcome now. Yeah.

I like it a lot. So there's speed, but what about the cost? Side of things? Because, uh, a market study, I don't know, might cost a couple millions, right? And return results, uh, I don't know, 3, 6, 12 months down the line, and then it's already too late. What about your cost and your, uh, time to market?

Yeah, um, I mean, it depends on the vertical, right? And yeah, over the course of the year per product, I'd say that's pretty, pretty accurate. We're, um, I guess consumption-based SaaS is probably the most accurate business model. People say AI as a service, people say SaaS nowadays. We're all trying to figure it out, right? But it's a subscription. It's gonna be in the 6-figure range, right? And depending on how many regions, consumers, products you're looking at, that changes the price. So it's like a node-based pricing model. Do you want to understand chocolate in Japan and beverage in the USA? Those are 2 different subscriptions. That gives you an idea.

Do you do worldwide or mostly US right now? And what about verticals? Is it mostly coffee, chocolate, or is it mostly wine right now?

Yeah, so we're working with global brands, right? They have products all over the world. Most of our presence is in the US, but, you know, we've done Singapore, we've done the UK, we've done parts of Europe. Asian countries we're moving into next. So it's happening. And then the other question you had was—

Yeah, you answered. Well, I mean the verticals, right? Chocolate or beer or—

Oh yeah. Yeah. So I'm just really excited to be moving beyond wine into soft drinks. Functional beverages are huge right now. You know, people aren't really drinking anymore. They want ashwagandha and their kombucha and and all those things. So we're very much about that. Beverage is huge. Um, next is beauty and personal care. Um, I think, really? Yeah, I mean, people would find it shocking how much more—

I didn't know that people ate their lipstick.

No, but there's billions of dollars being spent just to perfect skincare texture.

Like, yeah, and the smell as well, right? Like, and you also get into smell or just taste.

Smell, taste, texture.

Texture, huh? Texture. How— yeah, that's intriguing. How'd you do that?

It's like, um, it's just another— it's just another sensory human modality. So we have chemistry of products that have texture, and we have the preferences of consumers for those textures, and then we invent new textures.

And yeah, how would you describe the Magic Mouse texture? I love touching my Magic Mouse.

What is that?

My Magic Mouse. Oh, that's a good one, right? You talked about Steve Jobs, so I brought it up.

Words are not enough to describe the Magic Mouse texture. I can only tell you how good it is.

It's like a smooth surface and it's white and it's very minimalistic. Not sure if Steve was, uh, behind that one. Uh, talking about, uh, taste, right? What's trending coffee right now? What's trending wine? What about kombucha? Tell me, like, what are the trends in the US right now?

Uh, so chilled reds in wine, I'm really excited.

Chilled reds?

Yes, cold reds and, and sparkling reds.

Oh my God, isn't that a sin?

Uh, no, like Europeans, Russians, you know, Ukrainians, they've been doing that. Moldovans have been doing that for a very long time, and it's really good. And if you go to New York It's like buy the glass in every restaurant. Um, I think people should try it. Yeah.

What about coffee?

Coffee? I don't know.

Take Nespresso, right, as a case study. Like, I have probably, I don't know, 15 types of Nespresso next to me, right? Um, wow. I'm not too sugary, but I know on their website they sell a lot of sweet tastes. Um, yeah. And could you help them, for example, develop something?

Absolutely. Also, um, I think in coffee it's sweet if it has esters, if we're talking about the natural components. So it could taste sweet without any sugar, something a lot of companies are interested in, um, offering right now.

Yeah, I hate it personally. Like, uh, even Stevia, I got tired of it. And these sweet tastes, they— my brain goes like, uh, chemical, right? Like every time I'm— I have some of them, but it's like, I don't know, 5% of my Nespresso collection here. Maybe like end of the day for some reason, right? It's like 4 PM and it seems like less hard. It's like, oh, signal to my brain that I need to start chilling.

Absolutely.

What about decaf? Decaf is like notorious to taste like shit, right? Like, have you seen brands work on like a better decaf taste?

Yeah, decaf is like the non-alcoholic wine of coffee. It's just, it's real, it's so, it's so tricky, uh, because that That bite comes from the caffeine, and finding a replacement ingredient to give you that bite is really hard.

I have a potential client for you that I can refer you to as well. Do you know Go Brewing? I'll drop them in the chat here. Go Brewing.

Go Brewing, not personal.

Yeah, they do a bunch of non-alcoholic drinks, right? Um, and apparently they got the taste right. Um, But yeah, during my days, I mean, I don't drink often, and if I'm gonna drink, it's gonna be something with alcohol, I guess, right? Because I don't like the taste of beer enough. Um, have some companies cracked the taste side of non-alcoholic drinks?

Yes, uh, very much so. And I'm under a lot of NDAs, you can imagine. No, no one wants us to talk about how, uh, they've used AI to improve their products. They love to talk about the marketing part, but it is, uh, kind of a sensitive area right now. But yes, it is possible, it has been done, and the ones who have done it have been successful. That's as much as I can say.

So what do consumers want in that area? Is it, uh, reproducing the taste of alcohol and the feeling of alcohol, right, without having alcohol inside the drink?

This isn't directly answering your question, but what I will say is, is if you do a test on a non-alcoholic wine, something really interesting happens with consumers. The second you tell them it's wine, they think it's terrible. If you tell them it's anything else, just a beverage, they very much like it. So part of the problem is there's a certain expectation when you call it something.

So on a branding side, not— so what do you call it, right? Like, um, I don't know, brands like Mamida and all of that, uh, well, I mean, did they invent something? Isn't that called like a seltzer?

Like, uh, carbonation helps. Carbonation helps a lot because it gives you that bite that alcohol, um, usually offers.

About kombuchas, uh, I'm a kombucha drinker. I love kombucha. Um, And I've been to kombucha for like 10 years or something like that. And there's brands in Canada that I like a lot, like Gutsy. They cracked the taste code for sure.

Yeah.

What's trendy in that market?

What's trending? Oh gosh, I don't know. Like, as far as flavors and kombucha, I'll have to get back to you on that one. In Canada, it's usually like if I ask for an answer, for something like that, it's going to give me, um, a handful of answers, just so you know. Um, as in like 11.2% of the population is, is most likely to like this product or this product that doesn't exist. And then it's like a Venn diagram. So it is not the most satisfying answer, but is the accurate one.

Yeah, my approach would be like take the what the market wants, right? Let's say Red Bull. Let's say we're in the energy drink market and try to reproduce that taste, but maybe elsewhere in the market, in the alcoholic or non-alcoholic or low-alcoholic, uh, side of things. That would be my PMF brain in action if I would launch a company there, which I would never, because B2C is hell.

Yeah, what we usually do— I avoid doing copycats because it's not the best way. What I usually do is I will say, these are the consumers that have the palate for Red Bull. I want to go after those consumers, but the product can taste completely different, but the same consumers would like it. That's how you get market share.

Does the product taste—

way of looking at it. Yeah.

Does a product taste different if the marketing is different? Like, Red Bull is almost immediately associated with extreme shit, right? And I'm an extreme athlete and CEO and founder, whatever.

Uh, is that— is—

does that count into the play and the advice you give?

Yes, I mean, it definitely— marketing is a massive influencer, right? It's not just about what's in the product, right? There are all these other things like the celebrity endorsing it. Um, I'm not denying that, but what I will say is, is that's not the secret to making something like this work. The secret is this principle that people don't like what they don't like more than they like what they like. So it's not about controlling latent variables like that. It's about mitigating the risk of having someone have a negative experience because people remember the negative aspect of what they're experiencing more than the positive aspect. And if you can crack that, then you're going to be successful. Um, yeah, I, I don't know if that makes sense, but like, do you, do you hate a certain food like tomatoes or blue cheese or something you really don't like?

Not really, but yeah, blue cheese, uh, smells disgusting and just tastes weird. I feel there's very few brands that crack the exact taste and texture of blue cheese, right? Like the, the commercial ones, they all suck, but like the niche ones, I think they're very good.

Yeah, but imagine the best day of your life with the best possible marketing for a blue cheese. Do you suddenly like blue cheese, or is it kind of something you can—

yeah, for sure.

Like, that's the— that's kind of like part of the secret sauce.

Yeah, yeah. For example, I don't like champagne for some reason, even like the very expensive brands. I think one brand that's been really successful as of late has been Mateína. I'm not sure if you studied them, but like they've used all kinds of influencers including, uh, Huberman and all of that, um, and it, it tastes pretty good, right? And it's a stimulant, it's yerba mate basically. So they took like, uh, a herb that's been very popular in, uh, South America and Argentina, precisely in Brazil, and they commercialized rise it up there. I think that was a pretty good play.

Yeah, yerba mate has been around for a long time. It's almost like there's a resurgence now, like it's a recycled trend or something.

Who do you study nowadays, like CEOs or brands?

Who do I study?

Yes.

Well, I mean, obviously I keep up with all the big 5 in the current events, but more niche. I'm really interested— well, I'm good friends with like Rob May. I consider him an AI luminary. He's coming out with a book soon and he really thinks about AI from the business side and like second order effects that happen from implementing AI in workflows, things like that. Like the pragmatic part of making it work, which isn't always about the technology, right? It's about how does it actually function in an ecosystem? Are you creating bottlenecks that are gonna have to be alleviated in this workflow? Things like that. That's really important. The companies that figure out how to do that are the ones that are going to be successful.

And why do you quote Winston Churchill so much? Through how many layers of hell have you been through?

There are so many things in life, and I am probably not the first person to say this, but there are so many things that are easier than this. Um, it is, you know, running a business even in the best of times can be a nightmare. Um, it's really, really hard. And I know people go on podcasts and they always say, yes, it's really hard, it's so hard, but it really— it, it's, it's a crazy endurance test. And in some ways it's like the Navy SEALs psychologically, right? You kind of, you fail by quitting and you constantly shock yourself at how much pain and suffering you're capable of handling.

Um, over time, you know, like when I ask you the question at the start why I like black coffee so much, it's like that— it's the pain of it. Yes, like me, that's my mantra, right? It's like grit and pain, and that's, that's like most of my operating system. And I think that's why I like hard stuff. If I'm gonna snack on something as well, it's gonna be like, uh dark chocolate, right? Like very strong stuff that gives energy, very practical stuff, right? But yeah, I can relate to it, especially on your side, right? Like, have you raised capital for that one?

Yeah, we've raised $14 million to date.

You had to, right? Because it's like such a hard niche. How's it scaling so far? Like, um, do you feel you're through, like, most of the layers of hell? Do you feel you're approaching, like, Middle Earth?

Yeah, I mean, long ramp up, right? We should— for— I know $14 million sounds like a lot, but it's really a tiny amount of money compared to companies who have tried this and failed on just the technology side. Yes. I don't think we could have done it. I like— we wouldn't have been able to do it for 10 times the money. Honestly, it's a really hard problem. Um, the infrastructure of the company for a very long time was very tech heavy. And now we're, we're expanding into the business side and now our infrastructure is adjusting to that. So it's kind of like slow, painful ramp up for a long time and then everything all at once. And that's dangerous and chaotic and exciting and yeah, has its own set of opportunities and problems.

Yeah. $14 mil. Like when you think of companies like Coca-Cola or Lay's, they must have spent at least a couple billions trying to crack and figure it out, right?

Yeah, it's crazy.

Um, do you study these big boys? Like, uh, are they in your repertoire of companies you reverse engineer and try to understand and see where you can fit in their Tetris game?

We're talking to a lot of the Coca-Colas of the world. Um, I've talked to a lot of very intelligent people running R&D departments product development departments. We're in conversations and working with a lot of them. Um, some of this, I wish we had maybe had this talk, um, a little bit later because some of this will be made public eventually. We are unfortunately not one of those startups where we do a bunch of press releases with Coca-Cola or anything like that. We're doing really important work behind the scenes, and it's a fight to make it public. But, um, yes, Yes to your question. Yeah.

Uh, GTM— how do you approach GTM? You have like a salesperson? How do you get, uh, these guys in your pipeline?

We have 3 people in the company, myself included, who sell primarily to enterprise right now. Um, you know, we, we have salespeople, but this is still— we're still in that process where the founder, um, unfortunately is an important part of the discussion. And part of that has to do with is, you know, yes, we have software, but if you're going to have a scalable, repeatable subscription model, that means you already have to be in this workflow. And we're disrupting the workflow, like we're another step in the process, we're changing the process. So yeah, it's, it's pretty hands-on for a little bit.

What about, um, why don't you start a podcast targeting like CIOs at these big food and bev companies? You get, you interview them, and on back and you pitch them on your thing?

I think if, uh, I didn't have to sign so many NDAs, that would be a great idea. I think, um, I only— I so very much wish, um, I could share the crazy—

some of them will come on and if you have them sign something before, they'll just talk, you know, and it's like free publicity for them. But yeah, let's, uh, let's revisit that. You raised $14 mil as well. You're raising some more, like, is also that coming under the radar pretty soon?

I'm, uh, yeah, I'm, I'm looking at that right now. Yes, it is, it is. Raising is part of the plan. I think you're supposed to say raising is always part of the plan. Um, yeah, um, you're supposed to say that, you're supposed to say that, right? Um, you're always raising, or, or maybe it's the opposite, actually.

Yeah, I think it's the opposite.

You're not raising right now. I'll let you know when I'm raising.

Um, Yeah, there you go. Kat, where can people find out more about you and Tastree?

So tastree.com. It's like pastry but with a T. Should I give out my email? It's so sketchy nowadays. I'll do it. I'll do it. It's kat@tastree.com.

We'll scrape it and we'll reach out. But no, thanks for coming on. I love the honesty and yeah, it's pretty special what you've built. So that was Captain Charles Cormier for another