MLOps Zoomcamp: Free MLOps course. Register here!


Launching a Startup: From Idea to First Hire

Season 4, episode 7 of the DataTalks.Club podcast with Carmine Paolino


  • The Mom Test by Rob Fitzpatrick
  • Design Thinking by Robert Curedale


Did you like this episode? Check other episodes of the podcast, and register for new events.


Alexey: This week, we'll talk about building a startup as a technical person. And we have a special guest today, Carmine. Carmine and I are ex-colleagues. Carmine worked with me at OLX as a senior data scientist. At some point, he left OLX, he joined entrepreneur first, which is a startup factory, so to say. Now he's a co-founder and the CTO at FreshFlow. Welcome. (1:46)

Carmine’s background

Carmine: Thank you, Alexey. I'm glad that you could invite me to see you again. Great introduction. Just to get to know me a little bit better. I started programming when I was a kid at the age of five. I continued this passion my entire life. I went to university. I felt quite fortunate that I could actually know what I wanted to do. I know that a lot of people struggle with that. I did a computer science bachelor at the University of Bologna. I did some gigs as a freelance web developer for around a year. Then I published my thesis in a book. The name of the thesis was "large scale social network analysis". It was about how to parallelize social network analysis algorithms using MPI. A lot of C++. I'm glad I don't have to do that again. But it was really fun. If you want to check it out. It's published in Springer. (2:16)

Carmine: Then I joined my master’s in the Free University of Amsterdam, which was in artificial intelligence. Since then I fell in love with it and continue doing data science for five years, including at OLX, at another consultancy before, and at another startup before. At OLX I went to OLX-squared, which is the moonshot team of OLX, where you can innovate as much as you want. After that, I joined the data science for social goods in the Alan Turing Institute, which is a research institution in the UK for data science and machine learning. I mentored some students there. We helped the World Bank and Ofsted (Office of standards for earlier childcare in the UK) with some really exciting data analytics. After that, I decided to join entrepreneur first. (3:27)

Alexey: You said you started programming as a five-year-old kid. My kid is now five years old, so maybe you'll tell me how to teach him a bit of programming at some other time. (4:31)

Carmine: He has to be really not excited about watching TV. (4:51)

Alexey: Interesting. That's not the case. (4:57)

Carmine’s startup FreshFlow

Alexey: So tell us a bit about your startup. What do you do there? (5:06)

Carmine: I'm the co-founder and the CTO of FreshFlow. FreshFlow is a complete ordering system for supermarkets. We solve the enormous food waste problem that we have in the world. And also make more money for supermarkets and make more money for us. We have this really nice triple bottom line, which makes us a profitable business — not yet, but we will in the future, hopefully — and also will make more money for for the supermarkets and benefit the planet. (5:09)

Carmine: Avik, my co-founder, saw a problem when trying to sell to supermarkets another computer-vision based solution. He saw that they have a bigger problem, especially for fresh products. Fresh products have varying shelf lives. Even the same type of product like mangoes, one batch could have a week left, and another batch could have two days left. And it's really, really hard to predict that, predict how many you need, and at what time. We found that most supermarkets still use people on the floor, so-called "fresh product managers". Every single day, they spend two hours looking at the shelves, and looking at what they have in the back to figure out what they have to order. When we started our first pilot with Edeka, which is the largest German supermarket, with one store right now. And we are signing with other stores in the future. (5:46)

Doing user research

Carmine: We went there with user research. We learned that at OLX-squared, it was a great experience there working with the designers from Koos. We saw that the fresh product people go there at 6am in the morning. They have a lot of boxes to move around — they put all the things in place, which takes a couple of hours. Then they start the ordering process for the next day. They calculate what is going to be the demand the next day. They have to think about it in their head. They have to think about weather, seasonality, is there a big event in the town. In the specific town where we're doing the pilot, there is a big ramp to go to the to the supermarket. If that ramp is full of snow, nobody will go to the supermarket. They have all of these things in their heads for 200 products. It's insane. It is really an incredible work. I'm amazed it works at all. They're doing a great job. But we are humans, we're not calculators. So we create ordering system, that lives in an iPad app, for all they know. It lives mostly in the cloud, to be honest. It uses machine learning and reinforcement learning to figure out the perfect order. (7:13)

Alexey: This position, fresh product manager, that's a different kind of product manager. They actually manage products. (8:55)

Carmine: Yeah, it's a lot less about politics and a lot more about how many strawberries you have to order. (9:09)

Alexey: So, every store in every supermarket have this position? They have to wake up at 6am in the morning. They have to do go around, look at the shelves and see "These mangoes are getting spoiled. We need to do something about them." They do this every day and it's a lot of manual work. (9:17)

Carmine: It's very tough work. We were amazed that these ladies could do it that early in the morning and also keep fresh to think about all these other factors. Honestly, we were struggling to follow them, they were so fast. It was really amazing. (9:49)

Alexey: So you were sitting with them, watching how they work, and then try to replicate it? (10:11)

Design thinking

Carmine: Yeah, we were shadowing them for a couple of days. We went there for a week. This is how we work in general. It's called "design thinking". From my experience with OLX-squared, and with the book "Design Thinking" by Curedale — highly recommend that by the way — it's very much appreciated by our clients that we do that. We are able to really figure out what are the processes, how things go in the store, and then deliver the solution that they actually need, instead of what we think they need. This is big learning that I had from that experience with OLX-squared. Also, EF push another book, which I also recommend, it's called "The Mom Test". I would say it's a much lighter version of "Design Thinking". It's mostly about how to do sales calls and stuff like that. But it does help you figure out what to ask. You want to ask things that actually make the user tell you all their processes, and not tell you about the solution. Because there's always a difference between what they want and what they need. (10:17)

Entrepreneur first

Alexey: I don't think you just woke up one day and realized "Okay, there is this problem. Let's solve it". I'm curious, how did you actually come up with this idea? What did the process look like? (11:47)

Carmine: It was very interesting. I joined entrepreneur first because I wanted to have a startup. I wanted to do it since my bachelor’s. I had a few ideas. I tried to ideate with some people and come up with some ideas for a startup, but it didn't really materialise. The problem is that you're always so busy. We were trying to do it also having a job at the same time. It was really hard. You need to really dedicate your life to find a good idea — unless you have a good idea already. Also finding a co-founder is really hard. They need to have the same amount of time and the same life circumstances. (12:07)

Carmine: When I received the call from entrepreneur first, I didn't know about them. This was around February 2020 when I was in OLX-squared. I already got the position at the Alan Turing Institute. I was really keen on first doing these two things, because they were so cool. They told me that another cohort would start in September. That was exactly in line with with the rest of my timeline. I reapplied for the September cohort got in. I had an amazing time. Entrepreneur first's tagline is "the world's leading talent investor". They call themselves "talent investor", because they take you in because what you did in the past is quite interesting, the growth that you have is pretty big, or you know something specific that nobody else knows. They take a bunch of people like that, they give you a bunch of talks on how to do sales, how to do marketing. (13:16)

Alexey: This is something that technical people like you and me don't know. We don't know what marketing is. I have no idea what marketing people do. It's so nice that they also teach you these things. (14:44)

Carmine: Absolutely. They have experts in entrepreneur first. For example, for sales, or people that exited companies for many, many times. There are Venture Partners there because essentially they are venture capital for pre-seed companies. But they operate quite differently. So they have experienced venture partners, and they talk about these topics. They were even talks by Ben Evans, who is one of the best people in the VC space, used to be at Andreessen Horowitz as well. He's now a global venture partner there, and many other people that are perhaps less known, but really, really good. (15:04)

Finding co-founders: the “expertize edges” framework

Alexey: So, they put you through this training, but I guess that's not the only thing they do, right? (15:55)

Carmine: Of course. They assist you with a framework on how to find a co-founder. First of all, you will know if you're a CTO type or a CEO type, even though there may be some exceptions. Usually, if you have a technical background, you're a CTO type. If you have more business-minded background or domain edge, they call it a CEO type. (16:01)

Alexey: A data scientist would be a CTO and a product manager would be CEO? (16:35)

Carmine: Yeah, roughly. It could also change. They have so-called "edges", and you fit in one or two edges. The edges are catalyst edge, domain edge, and technical edge. The catalyst edge is further defined as catalysts doers and catalyst talkers. The catalyst doers are very energetic. They can make almost anything. They don't have a specific technical edge that is so sharp that only 10 people in the world have. But they're more widespread. Data scientists will be full-stack people. We, for example, were catalysts doers for EF. Catalyst talkers are flexible business people. They are the same thing but on the business side. (16:41)

Carmine: There are also domain edges. They are people that have in-depth knowledge of a specific domain. One of the co-founders had deep expertise in the pharma space. So he worked. He was General Manager at a company which got acquired by AstraZeneca. He worked there for 20 years. He's an absolute master of computer vision and pharma. That's a definite domain edge. Tech edges are people that have very specific skills in some tech. For example, there was a guy that did his PhD thesis on 3d printing human hearts. I don't know how many people do that. (17:49)

Alexey: It's as specific as it can get. (18:46)

Carmine: Absolutely. They give you this framework in order to start the conversation with the people. If you're domain edge, you will definitely talk about your domain. If you're tech end, you talk about that specific thing. If you're a catalyst, it's a bit more fluffy. You talk about what you're a most expert at and see what are the inter-weaving threads that you can find with other people. They are very interesting conversations. It's really like one of the best times ever. You can talk about crazy things from one day to another, you can have 10 conversations in one day where they span from "how to solve the climate issue" to "how to make a dashboard for FinTech companies". And there could be anything in between. You also go with your own gut feeling, trying to see who you will fit with, what kind of ideas are you most drawn to. You, of course, bring your ideas and you test them a little bit. (18:50)

Carmine: They have mentors who have a call with you every week. There are two mentors, actually. One is the relationship mentor. You can ask about "how can I approach people" or "there's something wrong in our relationship with my current co-founder, how can we fix it". You get the other mentor only if you have a team. They are quite experienced in the startup space. They will tell if what are you doing good, or what are you doing wrong in terms of market sizing. They help figure out if the idea has any chance of working. Every single week they give you a score between one and four. It tells you how you're going to do on the final examination. It's called the "Investment Committee", which happens in January. (20:13)

The structure of EF program

Carmine: This entire thing starts in September. In January, you have investment committee. EF program is divided in few phases. The first phase is the form phase, which takes eight weeks. Here you try to find your co-founder. After the form phase, you have the company building phase: you have to prepare all the material for the Investment Committee. The Investment Committee is a bunch of venture partners from EF. We actually had Benedict Evans in our committee, amongst others, so it was a bit intimidating. It was very, very helpful and very, very useful for us in the future. Fortunately, we passed. (21:23)

Carmine: That was my previous co-founder. I had a few teams in EF. My current co-founder, THE co-founder, is a machine learning scientist with a similar background, but much younger. He's 26. He just graduated from ETH Zurich. He was quite active in the entrepreneur space. This is now his third company. It's quite insane at 26 (22:06)

Alexey: They decide if they want to invest or not, right? That's the final examination. (22:15)

Carmine: That's the final examination. (22:19)

Alexey: If they want to invest, then you get the money. If they don't, then... (22:21)

Carmine: Then bye, bye. (22:26)

Alexey: I've heard some people still try to pursue the idea, to find other ways and find other investors. (22:29)

Carmine: Yeah. People that didn't get the investments, got other grants, or were incubated in other incubators, or went for competitors of EF. In the end, if you want it, you can do it. That's, that's a big lesson. (22:38)

Alexey: You mentioned that your co-founder has experience in pharma. (22:58)

Carmine: That was my previous co-founder. I had a few teams in EF. My current co-founder, THE co-founder, is a machine learning scientists with a similar background, but much younger. He's 26. He just graduated from ETH Zurich. He was quite active in the entrepreneur space. This is now his third company. It's quite insane at 26. (22:06)

Alexey: I don't know what's more insane — programming at five or that. (23:06)

Carmine: Thanks, and I hope that these credentials will come up in our products. He's very impressive in that sense. He founded a company in motion analysis, then another one in cryptocurrency, in China, while he was studying. And now FreshFlow. He's the CEO, he's only on the business side. And we decided that just because it fits more with what our backgrounds are. I have longer technical experience, and he has longer entrepreneurial experience. It works. And it's great to have a technical co-founder. We were an exception. That's why I said "spoiler alert" before. (23:48)

Coming up with the idea

Alexey: How, when, and why this idea of your business came to your mind? From what I understood about your co-founder, he's doesn't strike me as a person who spent his life overseeing a supermarket. How did you actually get this idea? (24:47)

Carmine: As I said before, he was developing a computer vision recognition product during his thesis. He started to sell it to different supermarkets, he started to have calls with different supermarkets, being entrepreneurial-minded as he is. During these calls, he realised that there are a lot of problems in how these supermarkets order food. So something was already on his mind. At the same time, when I broke up with my team, the pharma guy, he also broke up with his teammate at the time. We found each other the last week before the closing of the form phase. We really wanted to have a startup. (25:09)

Carmine: We were friends from the beginning with Avik. We were always having beers. We were also trying to ideate with other people, obviously. You want to stick to what EF people say — try to find a business person that started from business. We were talking with a couple of people, but we found that we were ideating much faster with each other. We started ideating about product recognition — if we could use what he's done in his thesis. We started thinking, "We could do something for pricing. We can recognise if products are going bad". I was thinking also in a "social good" sense because it's dear to my heart to do something about climate change. It was a big topic between my projects in EF. In the end, we started to see that there are a lot of problems with estimating shelf lives and keeping track of what is going bad. (26:15)

Carmine: We started ideating in that sense. "Can we do something with computer vision that looks at fresh products, and tells you if they are going bad?" We thought, "maybe we can do an app?" With this idea, we went to the so-called "super check-in", which is like a mini Investment Committee. So, it's an app that you can use as a consumer. You can go in the store with it, take a photo of a fresh product. For a packaged product, you can take a photo of the barcode and the expiration dates. But for fresh products, we will do the analysis of the ripeness in the app using computer vision. Then it will give you a discount to apply at checkout. That didn't fly very well with the Investment Committee. (27:28)

Carmine: So we went back to the drawing board. We learned from that. First of all, it was hard for people to get another app just to get discounts. It's not enough to have just discounts. Then we learned that it's not that easy to figure out the shelf life of fresh products just using computer vision. So we pivoted. We started thinking and realised, "Oh wait, I remember this conversation that I had with these people. I remember that wasted stockouts are a huge problem." We wanted to avoid all that because our edges are in computer vision — we did a lot of computer vision. But we saw that this could be a much bigger problem. Then we started looking at the numbers about food waste. And it's completely mind-boggling! Just to give you an idea. Food waste is actually accounting for six times more carbon emissions per year, than the entire aviation industry. (28:34)

Alexey: That's a lot. (29:55)

Carmine: It's insane. When I read that, I was like, "I can believe this. This is crazy." We're wasting $1.7 trillion per year on food that never gets sold. But someone has to buy it. The supermarkets have to buy it. And out of all food waste, supermarkets contribute to 10% of that. So if we can contribute to 10% of six times the aviation industry, we're almost fixing the aviation industry. (29:56)

Carmine: We got really, really excited. Then we saw that there are some exceptional companies in the US that are doing something similar. Some of them are Shelf Engine and All Fresh. They are doing well. And we really, really appreciate what they're doing. We wanted to follow suit and started our own company FreshFlow here. (30:35)

How important is going through a startup accelerator?

Alexey: Nice. A question from Anika. How important is to go through a startup accelerator? Would you be able to do this without entrepreneur first? (30:59)

Carmine: No. As simple as that. You need to be extremely lucky to have the kind of connections that entrepreneur first gives you. Maybe go to a very specific place. For example, in Avik's case, ETH helped a lot because they have an entrepreneur club. If you're interested in that, you join the club, and you see a lot of people with ideas and professors that can help. People actually create their startups during university time. If you're in the Ivy League, there are a lot of people who start startups. But there was no such thing in Free University of Amsterdam and University of Bologna. I didn't find anything like that. Having these kinds of possibilities changes the game completely. I have a lot of friends now from entrepreneur first, just because we vibe so much together. We're also ambitious and creative and risk-taking. (31:11)

Alexey: So it's about building connections, it's about networking, right? If you have already a strong network, from your university, then you don't need it. I know that in Berlin there are startup meetups or founder dating meetups. It's possible to build this network. But entrepreneur first helps you do that. (32:26)

Carmine: Yeah. I would say it's still hard. Even though you would have the network, then you need to find a co-founder that is really driving with you. You need to have the time to think about the idea and really develop it — and try to do it as quickly as possible. That's always the conundrum of being in a startup. You want to always be one step ahead of you are. That requires a lot of time. I'd say without entrepreneur first it would have been really, really hard. (32:53)

Finding your first client

Alexey: How did you find your first client? Running a pilot with Edeka is awesome. Is it your first client? (33:24)

Carmine: Technically, we had also another client, which came from Avik's previous dwellings in the supermarket space. This other client is actually Volg, which is the third-largest supermarket chain in Switzerland. In Volg we tried a slightly different product. We did a pilot, and it went well. For the first client in Germany, Edeka, we entered through the Edeka foodtech campus, which is a startup accelerator by Edeka. They also give you mentorship and they connect you to the right Edeka stores that are perhaps more open-minded to new things. After explaining our idea, we got in. We connected to the right people in Edeka and we refined our idea and our product. And then we started working with Edeka Cassius. (33:33)

Alexey: Was it before you graduated from entrepreneur first? Or after? (34:39)

Carmine: After. (34:44)

Alexey: So you don't need to get a client to graduate, right? (34:45)

Carmine: Well, it'd be beneficial. But they bet on us. First of all, we formed the last week of form phase. We had only four weeks to get our narrative together, get the pitch deck, get all of the things together — all of the possibilities for the future and the business model. Quite a lot of work in a short amount of time. After that, you start building. We started building the algorithms and other things at the same time. It was quite busy anyway. (34:50)

Carmine: But you may need to show that you have some traction early on. We already had some traction at the time: we had a few leads with other supermarket chains in Italy, and in Germany. It didn't go through at the end, but we had around 4-5 leads. There needs to be some traction for sure. (35:30)

Alexey: How long did it take to get the first client? The time between coming up with the idea and securing the first client. (35:56)

Carmine: Unfortunately, the supermarket space is not that fast. The sales cycle is quite long. Our contract with our first Edeka store was signed in April. And we graduated from EF in January. We officially registered as a company also in April. It just takes time to do that. There's a lot of paperwork here in Germany. (36:07)

Finding investors

Alexey: I also prepared a question about investors. I guess you answered that. So entrepreneur first acts as an investor. There is a board and they decide whether they should invest in you. Is it the only investor you have? (36:40)

Carmine: Ish. There are things that we will announce later. But at the moment, officially, there is only them. We're lining up angel investors at the moment. And we have an amazing guy as an angel investor, which we'd love to announce, but it's in the works. (36:57)

Alexey: You don't have to announce now, but maybe you can tell us about the process. How do you actually do this? With an entrepreneur firstly, this committee decides if they invest or not. But once you don't have the support from EF, how do you go about finding an investor? (37:20)

Carmine: EF gives you a fundraising manager after you graduate. They help you with structuring the narrative to find the investors. They also give you a platform called "Demo Day", which, in Corona times, is just a web page with some YouTube videos. It used to be that you go to a theatre and give a pitch in front of 100 invited investors. (37:39)

Carmine: We decided to not go for them. We found that we need to be proven with the numbers and the measures that we get from the pilots in order to gain traction since it's such a slow industry. Also, it's very hard to sell to supermarkets. Because imagine you're substituting the core parts of the entire ordering system of the supermarket. Think about what a supermarket does: you get stuff in and you get stuff out. You have to present them well. So we fix the "get stuff in". And then they have to think about "present them well" and "get stuff out". It's pretty hard to sell. And we want to raise on our terms later. (38:10)

Alexey: What are the places where people can look for investors? Are there meetups? (39:16)

Carmine: Well, investors approached you actually. (39:22)

Alexey: How do they find you? Do you just post on LinkedIn "Hey, I'm looking for investors"? (39:26)

Carmine: They search for you on LinkedIn. Just by the fact that you have a "co-founder" in the name. Also, they may have bots that parse the register of companies in Germany. The VC firms know all the companies. They have a bunch of analysts who go through all these companies. They look them up, they see what they're doing. If they think it's good enough, they will shout an email. (39:32)

Carmine: It's not that hard to find an investor. But finding good investors is up to you. There are investors that you want to have on your board. They have specific expertise. For example, investors in the foodtech industry are great for us. Other high-profile investors are usually less hands-on, they care rather than control. You need to do a little bit of the work to get them, perhaps approach them yourself. (40:07)

Consequences of having a bad investor

Alexey: You said that there are just investors and there are good investors. What could be the consequences of going with a bad investor? (40:13)

Carmine: You may get a bad term sheet, where they take a lot of your equity for not so much money. You may get people on the board who are trying to control too much. There's not so much trust, but it's rather a culture of "We give you the money. We want the money back immediately. Grow as fast as possible. Blah, blah." It's common that investors want to do that. It's their job, ultimately. But some of them do that in an empathetic, nurturing way that is more enjoyable for everybody. And some do it in a more predatory way. At the moment, we don't have anybody on the board, because EF gave us a convertible note. So it's very, very chill and very quiet. Maybe Avik, my co-founder, would have more insights into this question. (41:04)

Splitting responsibilities between co-founders

Alexey: You have pretty similar profiles. Even though he's more entrepreneurial, both of you are engineers. How do you split your responsibilities? What do you do? What does he do? (42:24)

Carmine: At first, we wanted to do everything together. But then we decided it was not going to help much because we'd have to synchronise ideas on every single little thing. And the productivity gains aren't much higher. But it's good at first to think about the major things together. (42:43)

Carmine: But now, I'm fully the CTO. I do all the tech except the mobile app. He does all of the sales and business development: the outreach to possible clients, talking to them, talking to investors, making the pitch deck. I do all the pipelines, the code, data mangling, the dashboards, connections to the app, machine learning algorithms, and experiments. All the engineering stuff. (43:08)


Alexey: Do you have other people in the company now? (43:47)

Carmine: Yeah, we do. We had the mobile developer, a freelancer. He created the app. Amazing guy. Unfortunately, his contract ended. But we hope that in the future we can work together. He did an amazing job. We also have a working student from the University of Mannheim. He helps Avik with the business side. We're going to start hiring more people soon. (43:50)

The importance of delegating

Alexey: Since the topic of hiring is already on your mind. How do you decide who to hire first? You needed an app, you hired a freelancer. But for a permanent employee, how do you decide who to hire first? (44:33)

Carmine: It's the same thing for the app and for the permanent person. It's very simple. When you hire, you know what you need. But a good principle is don't try to do everything yourself if you're not an expert in it. Ask for help. "Asking for help" in this case means hiring someone. At first, I thought I could pick up Flutter and do the app. But it would have pushed us several months back: I would need to learn a new language, then code the app in that language. I would probably do a worse job than what Jeremiah did. So, in the end, it doesn't really make sense. We had the investment already. So why not use it and spare my time to do something that I can do much better. (44:57)

Alexey: As a co-founder, you need to know how to delegate, right? You need to let things go and not try to do them yourself. I am pretty sure you will do an app, but it will just take more time. (46:10)

Carmine: Exactly. You may want to be the one doing all of it. But it's not shameful at all to ask for help. And it's appreciated by the clients as well, they see the product faster. By the investors, because they see their money back faster. So, by everybody. (46:25)

Making work attractive to hires

Alexey: I don't know if you had this problem. But I imagine that you as a startup, the package you can offer to candidates is less attractive than what big established companies such as Zalando, Amazon or OLX can offer. How can you manage to make this offer package attractive? Have you thought about this? How to attract people to work at a startup? (46:53)

Carmine: We hired internationally. We were not afraid to go into different countries. And fortunately, our mobile dev joined because he loved the technology, and he wanted to do something in that technology. He used to do it in his free time. He felt that he can at least have some money out of it. (47:25)

Carmine: This is a special case of an extremely passionate person who really wanted to bring something to the world. And maybe also have fun and some money. In general, our offer was attractive, because we'd pay the same rates in all parts of the world. For example, for an Indian developer, that would be an amazing salary. (47:54)

Alexey: So for somebody from Berlin, it will be an okayish one. If they go to a bigger company, maybe they would offer more. But for somebody from a place where the average salary is lower than in Germany, for them is great. (48:22)

Carmine: COVID opened this amazing possibility of remote working to everybody. We just opened our eyes and we were like, "Wait, we can do this". So we just took advantage of that. (48:39)

Plans for the future

Alexey: Nice. What are your plans for the future? (48:52)

Carmine: More clients, more product, more shipping, more launches, more money, more investors more hires, bigger clients. And grow from there. This is of course the plan. But in the longer-term future, we want to be the grocery food retailer OS where it's a one-stop-shop for all of their logistics and supply chain needs. We also want to go in the other steps in the chain, So go up to the distribution centers, the fulfilment centers, and maybe go beyond that towards even the farmers. There's so much that can be done at all these levels to predict what kind of food needs to be on the shelves of the supermarkets. (48:56)

Just-in-time supply chain

Carmine: We can have so much impact by having one solution that would enable the just-in-time supply chain of Toyota. For example, I need a banana right now, because I'm hungry. My system will know, "it's Friday, it's 6pm, Carmine just had a long day and wants a banana". It then preemptively would predict that months in advance. The farmer will be like, "Okay, I need to produce this many bananas". (50:14)

Alexey: I think Amazon is doing that already. I sometimes feel like Amazon can read my mind. I ran out of shampoo recently. I went to buy it. And then, on the first page, it was already recommending me the shampoo. (50:50)

Carmine: Interesting. That's another side, the recommendation side, which is quite similar to forecasting. I learned a lot of things at OLX that come back here in this work. It would be much more towards forecasting. With the just-in-time supply chain, you have a lot less inventory space. That means less need for huge warehouses, less need for transportation. Trucking and all burn a lot of fossil fuels. That's not great for the environment. It's not just the problem of "the rotten food releases CO2 to the planet". (51:04)

Carmine: A lot in the supply chain can be fixed by ordering the exact amount. Ordering the exact amount is really, really hard. If you order too much, you're producing just waste. The retailers do that, because they don't want to have empty shelves. That's really bad. That means losing customers. The customers will go to another store if their shelves are empty. With fresh products, you have them in boxes. So if you have the last bananas or the last two bananas, it's basically like the shelf is empty. Because nobody takes the last two bananas. (51:54)

Alexey: Yeah, I know. I wouldn't. (52:36)

Carmine: Exactly. You think something is wrong with them. Given this consumer behaviour and other factors, they need to keep stocking the shelves. It's really hard to stock them to look full even at the end of the day and there's nothing in the backroom. We found that it's not the case. Because it's really tough. So we tried to fix that problem. (52:39)

What would you have done differently?

Alexey: Would you do anything differently now if you were starting again? (53:09)

Carmine: Yes. I would not use open-source infrastructure. I'd directly use GCP. We switched to it from Kubeflow, an open-source MLOps infrastructure. It was fantastic when it worked. But then we upgraded it to 1.3, and it started breaking in different ways. Also, 1.2 gave us a lot of trouble with the installation. I spent one and a half months taking care of Kubeflow. For a startup, it's not a great idea. At the end, I just said "fine, we use GCP". I'm glad we went for GCP. It has much better interfaces and software compared to AWS. (53:13)

Carmine: We're using BigQuery, Cloud Run, and Cloud Functions. It's an event-driven architecture. Everything that comes in gets pushed around in these pipelines. We use FireStore for the app API. It's lightning-fast. It's beautiful. And it just works. (54:21)

Alexey: So, for people who are starting a startup, you'd advise going with a managed solution, right? (54:42)

Carmine: Yeah. Go with a managed solution instead of an open-source solution. At least at first. Also, Google and AWS give startups some credits if you're invested by an accredited investor of them. We received 100k in credits from Google. It means that we can use their most expensive solutions, and sleep well at night. (54:51)

Alexey: So they bet on you, they bet that you will grow. That's why they invest 100k in their services in you. That's very nice. (55:28)

Carmine: Yeah. And in the end, if you like their services, you will continue. This is one of the things in the tech space that we could have done differently. (55:37)

Advice for people starting a startup

Alexey: Do you have any other advice for people who would like to start building a startup now? (55:51)

Carmine: Everybody says the same. "It's going to be hard. You're going to have tough times." It is true. Just be prepared. Be emotionally intelligent. Sometimes a sales call doesn't go well. It's not you, okay. It's not personal. Maybe sometimes you don't talk about the specifics of your solution. There's no need. Sometimes you can say just "IP". There could be a concern that you cannot address at the moment because you're just too small. You don't know certain things. (56:07)

Carmine: You struggle at first because you want to have the best clients — in order to have the best contracts — in order to have the best investors — in order to be able to hire the best people. But you're not there yet. And you're always trying to get to that point. It's really hard to cope with all the loads that you put on yourself. Be mindful of that. (56:52)

Don’t focus on skills only

Carmine: And, I would say, find a co-founder who's on the same page with you, somebody you can communicate really well with. They don't have to be the best salesperson in the world if they're on the business side. Or they don't have to be the best technical person in the world. It helps, but up to a certain percentage. (57:09)

Carmine: My girlfriend is a recruiter. She recently did a presentation that says that you should hire for 10% skills. The framework is, how much skills, behaviour, and motivation play out in the final performance of employees. The skills are about 10% to 20%, and the rest is motivation and behaviour. So, hire people for motivation and behaviour. If you get motivation, behaviour, and skills, you're golden. But you're going to speak every day with your co-founder. If he's motivated enough, he's going to figure it out. If his behaviour is great, he's going to figure it out as well. So that's another advice. Try to get people who don't fixate too much on skills. (57:50)

Getting motivation

Alexey: How do you get continuous motivation to do this stuff in a startup when things don't work as good as you expect? (59:14)

Carmine: This is kind of in you. I don't know if there is a way to get motivation. For example, I really care about climate change and the food waste problem. It seems to be the underdog of the climate change spectrum of problems. It's right there in front of us, and it is massive — six times the number of CO2 emissions of the aviation industry. If you think about food waste as a country, in terms of carbon emission it will be the third country in the world. After the US and China, there'll be food waste, and then it will be India with half of the carbon emission of food waste. That's just insane. It seems we have to do this. Also the opportunity space is pretty big. The food retail industry is now the process of digitalization, and the opportunity for companies like ours is great. And then you work with the right people. That's what you need. But motivation comes from the inside — from what you care about. (59:39)

Am I ready for a startup?

Alexey: The next question: "I'm 30 years old. I want to be a CTO in the startup that I'm launching with a CEO." I guess, it's pretty similar to what he did. His main focus is machine learning and knows little about front-end development, DevOps, and all these things. The question is, do you think it's enough to start as the CTO or not? (1:01:05)

Carmine: It could be tough sometimes. But if you use a managed solution, and you can spend some time learning new concepts, I don't think it's a problem that you have only the machine learning background. What I said before, the framework that they used in my girlfriend's company, 10%-20% skill, 50% motivation, and then the rest is behaviour. (1:01:35)

Alexey: Do you think he should learn something now before becoming a founder? (1:02:04)

Carmine: For sure. Much better now. Because then you have the pressures from investors, from clients, from your co-founder and your hires. So please, start as soon as possible. (1:02:08)

Alexey: But then you will never start. You will always think, "I'm not ready yet. I need to learn this thing. I have this new framework. There is this new book that everyone says it's good." How do you know that you already? (1:02:19)

Carmine: You don't. You will never feel ready. I have more than 10 years of experience in software engineering. And yet I didn't think I was ready for a startup. Until I joined the FN, I was like, "Wait a second, I'm actually not that bad. I can actually do this." You can also jump in and figure it out on the spot. (1:02:33)

Carmine: If you go for entrepreneur first. If you get in, start as early as possible. Not necessarily technical stuff. But things like figuring out the problem space. Start doing something if you can. But also don't be afraid of just jumping in. I actually think that just jumping in will propel you to know these things by itself. It's a bit of a balance. I know, this answer sucks. Because "what is it exactly that you're trying to recommend?" But either works if you have enough motivation. (1:03:01)

Importance of a business school

Alexey: What do you think about the importance of a business school? (1:03:45)

Carmine: Not important whatsoever. (1:03:50)

Alexey: You'll be fine without going to a business school and without an MBA. Right? (1:03:54)

Carmine: What I learned from other friends who did MBAs is that you don't need one for a startup. They teach concepts for running large-scale businesses. In a startup, you have a set of challenges that are specific to your markets. You don't need to know these grand schemes of things kind. You don't need to know a lot about a specific business case or a specific thing in a very structured way. It's is much more about shipping it as soon as possible. That's what you want to do. (1:04:02)

Alexey: So it's more about the mindset, right? (1:04:46)

Carmine: Yeah. (1:04:48)

Advice on finding a co-founder

Alexey: What happens if you don't find a co-founder? (1:05:07)

Carmine: Then you're out. (1:05:10)

Alexey: You have to have a co-founder? (1:05:15)

Carmine: You have to have a co-founder when you are in the form phase. They give eight weeks. Use them correctly. If you go for EF, try to get into a team as soon as possible. Don't waste your time thinking, "is this the right person?". Figure it out, get some answers, of course, don't get into it with the first person that you meet. But if you vibe enough, just go for it. That's kind of the startup mentality. Just go for it. (1:05:17)

Do I need EF if I already have an idea?

Alexey: Would you recommend starting with EF if you already have a clear idea? (1:05:57)

Carmine: I would actually start with EF even though you have a pre-existing idea. I also had a pre-existing idea. I actually tried it in the first few weeks of EF with another team. We figured out that the market wasn't ready for that idea. Being in EF will teach you how things work in the real world. Sometimes you're stuck in your own bubble. We worked at OLX, we may know a lot about marketplaces and data science. But not about the business side of things, like how investors think about who to invest in and so on. It will give you that kind of mentality. (1:06:09)

Having a prototype before the pitch

Alexey: The last question for today. Did you already have a prototype before your pitch? (1:07:06)

Carmine: For the first pitch, for Super Check-in, we had a demo for recognizing the ripeness on bananas only. It was web-based, with TensorFlow JS. But we didn't have a demo for the forecasting solution. (1:07:11)

Alexey: You just need to be able to pitch this idea, have this idea very clear, so that the investors can imagine how it will look like? (1:07:38)

Carmine: They want to see that the market is big enough. And that you're the right people for the job. And you can find clients. They usually trust you on the tech side if you have the credentials for it. They look at your CV, they're not going to make a technical interview. (1:07:51)

Alexey: That's good. We can put a few things more there. No, that's bad advice. (1:08:18)

Contacting Carmine

Alexey: So the last one. How do people find you? (1:08:25)

Carmine: I’m on Twitter @paolino. I'm also on LinkedIn, just search for my name Carmine Paolino. I have a website that I don't update — Also is our website for FreshFlow. If you want to email me, my personal email is If you want to email me for work, it's (1:08:29)

Alexey: Thank you very much for joining us today and for sharing your experience with us. Also thanks to everyone for watching, for being active, and for asking all these questions. (1:08:58)

Carmine: Thank you so much, Alexey. It was really nice. (1:09:08)

Alexey: Yeah, it was really great. Thanks a lot, everyone and have a great weekend. (1:09:10)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.

DataTalks.Club. Hosted on GitHub Pages. We use cookies.