What Europe’s Startup Gates Look Like From the Wrong Side

STATION F’s Landing Zone in Paris accepted Dastan Tomanbaev. Getting there would have cost roughly €5,000 — a plane ticket, accommodation, and often an IELTS certificate he’d have had to travel and pay to sit. He had none of that. So the invitation he’d have needed thousands of applicants’ luck to get simply expired, unused.

Every European accelerator, hackathon and fund likes to describe itself as open to talent “regardless of background.” Tomanbaev’s account of what that openness actually requires — a passport that grants easy travel, a few thousand euros in reserve, an English certificate, a registered company — is a useful check on that story, because he’s a case study in having the talent and none of the rest.

He’s the 18-year-old, self-taught founder of Solifon AI, building a browser-based gesture-control product called Air Canvas alone from a village in Kazakhstan. He was accepted to STATION F Paris and invited to ETHGlobal’s Tokyo hackathon and had to decline both, purely on cost. Below, in his own words, is what that gap between being accepted and being able to attend actually looks like — plus the rest of his story: training a model on a 4GB phone before he owned a laptop, and the speech-therapy product for children he ultimately shut down.

Part 1 — Where it started

Take me back to the beginning. What was the first thing you ever tried to build with AI, and what happened when you tried to train a model on your phone before you had a laptop?

“My path started in Year 9. I’d just begun learning to code — JavaScript first, which later became the foundation for my gesture-control system, Air Canvas. But before I ever had a computer, my first real attempts at machine learning happened on an ordinary smartphone with 4GB of RAM.

I’d got it into my head that I didn’t just want to use AI — I wanted to train my own model, right there on the phone.

Reality hit fast. Processes kept crashing from lack of memory; I had neither the processing power nor the storage. I couldn’t get far. But that hardware ceiling is exactly what pushed me to look for alternatives and go deep into web technologies and browser-based solutions — the kind that can run anywhere, regardless of how powerful the device is.”

The “Plus One” foundation gave you your first laptop. What did that moment actually feel like — walk me through that day.

It was pure, uncomplicated joy. It wasn’t just a gift to me — it was the first laptop I’d ever owned. Before that, all my potential had been locked inside the tiny screen of an underpowered phone.

When I opened that laptop, I felt like a door into the real world of technology had genuinely opened in front of me. That day I understood something important: I had no more excuses left. I now had the tool. Everything from that point depended only on how much effort I was willing to put into the code.

You’re self-taught through Google Colab and Kaggle. What’s the single hardest technical wall you hit, and how did you get through it?

The most frustrating, and honestly the harshest, wall was understanding how cloud infrastructure and memory actually work. As soon as I had a laptop, I threw myself into building my first dataset — by hand, piece by piece, first 1,000 examples, then up to 4,000 rows.

I uploaded everything to Google Colab, ran the training, waited for it to finish… and forgot to write the script that actually saved the model’s weights to disk.

The Colab session reset, and all that work simply vanished. It was a bitter but genuinely important lesson. I had to start again and write proper save architecture this time.

But the moment the retrained model finally generated a response, and correctly stated its own name and the name of its creator — that made up for every one of those lost hours.

Part 2 — Mother’s Heart

You’ve said your own childhood speech delay is the reason you built “Ananyn Zhuregi” — “Mother’s Heart.” Can you describe what that was like for you as a child — what you understood and couldn’t say?

I went through this myself as a child, so the problem was very personal to me, and I genuinely wanted to help children in the same position. But once I got into the machine-learning architecture, I ran straight into the harsh economics of AI.

To train a neural network to understand specific defects in children’s speech, you need datasets of millions of audio samples, enormous computing power, and budgets in the millions of dollars.

Even a company like Google spends colossal resources and years training audio models, and there’s still no 100% guarantee of a perfect result. I realised that enthusiasm alone wasn’t going to break through that technological wall.

How did the idea move from “I want to help kids like me” to an actual working product — what did the first version do, and how is that different from what you’re building now?

To be completely honest, the first version wasn’t a working AI product at all — it was a UI concept, a clickable prototype. There was no complex analysis under the hood, just a button to visualise the interface.

I built the concept to show my vision. But building a full medical-grade neural network is a task for a national healthcare system to solve, not for an 18-year-old solo developer from a village. Once I understood those limits, I changed direction completely.

I’m now building Air Canvas — browser-based gesture control — because it’s a product I can actually build, scale and control 100% by myself, today.

You described Mother’s Heart as offline-first, built for families in villages without stable internet or nearby specialists. What does “offline AI” actually mean technically, on a standard smartphone — what can it detect, and what does it deliberately not try to do?

Technically, ‘offline AI’ means running quantised models locally on the device’s own processor. I’ve had experience running models like this — LLaMA-format models, for example — locally on my laptop, and they work without an internet connection.

But with Ananyn Zhuregi, I understood that moving heavy audio analysis onto a cheap, low-power smartphone in a village was simply not achievable with the hardware available.

So the concept deliberately never tried to make ‘medical diagnoses’ — that would have been an ethical line I wasn’t willing to cross. Its theoretical job would only ever have been basic support until an internet connection became available.

Part 3 — The gap between building and accessing

You were accepted to STATION F’s Landing Zone in Paris and invited to ETHGlobal’s hackathon in Tokyo, and had to decline both — purely on cost. Describe the moment you realised you couldn’t go. What runs through your head in that moment?

It was frustrating, of course. You get the email that thousands of startup founders dream of, and then you realise your path is closed simply because you don’t physically have the money for a plane ticket. But I didn’t spiral into it, and I don’t regret it. I reasoned it through: life goes on — maybe neither I nor my product were actually ready for the global stage yet.

I just closed the invitation tab, opened my code editor, and kept working. If I can’t get to them in person, my code will still reach their screens.

What’s the actual number — roughly what would it have cost you to attend one of those, and how far off was that from what was possible for you?

I’d never been abroad, so this was a shock to me. The problem isn’t just plane tickets and accommodation — that’s a minimum of €5,000, an astronomical sum for my family.

The problem is also the passport and the requirements. Unlike Americans or Europeans, we don’t just get a visa — it’s a complicated bureaucratic process. On top of that, a lot of programmes and funds require official certificates like IELTS 6.0 or higher.

I have the knowledge, but I don’t have spare money or time to travel to a city and sit a paid exam for a piece of paper. The gap isn’t measured only in money — it’s measured in passports, too.

You’ve said many people don’t want to see an 18-year-old from a village where they expect “adult professionals” to be. Can you give me one concrete example of that — a specific conversation or rejection that stuck with you?

It happens constantly, especially with state programmes or when looking for work. I show them real, working projects with neural networks that I built myself. And they look at my age and my registered address and ask: ‘Where’s your official work experience? Do you have a diploma or a certificate?’

The system isn’t looking at what you can actually do — it’s checking whether you fit their standard template of an ‘adult employee.’ If you don’t have the paperwork, your real skill doesn’t exist to them.

What’s the most frustrating part of trying to get into programmes like Astana Hub or Alem AI without a registered company or investors behind you — walk me through what that process actually looks like from where you’re standing.

What’s most frustrating is that this ecosystem is often built for impressive reports, not for supporting real innovation at the very start. These incubators expect a ready-made businessman to walk in with a registered legal entity, starting capital and a city-based team already in place.

It’s like a closed club ‘for insiders.’ Innovation is often born in garages and villages, but the system demands you become a corporation on paper before it will give you a chance. It kills the spirit of building a startup and turns everything into endless bureaucracy.

Part 4 — Where things actually stand today

Solifon AI seems to have gone through several different shapes — a multi-model chat assistant with science simulations, then Mother’s Heart focused on speech development, and now Air Canvas and a broader AI workspace. Can you walk me through that evolution and why each shift happened?

The projects kept changing because my understanding of what a real business actually is kept changing. I started out running on enthusiasm, but I quickly realised that bolting an API onto a website isn’t innovation.

I kept searching for a niche where my code would have genuine value, until I arrived at the current model — subscriptions and more sophisticated interfaces like Air Canvas.

A few months ago you posted about exploring a full buyout of Solifon AI for $30,000, including the brand, code, dataset and press coverage. Where does that stand now — is the project still for sale, and how does that fit with where you’re taking it today?

I’ll be honest: at that point, the project wasn’t worth $30,000. It was just a website without a unique domain, wired up to a third-party language model’s API — something like Claude.

Anyone with a subscription and basic knowledge could have replicated it. There was no ‘killer feature,’ no protected technology in there. I wanted to sell it because I needed resources. I understand that now. So I stopped trying to sell thin air and started building something genuinely complex instead.

Is Mother’s Heart — the speech-development project — still active, on hold, or fully set aside in favour of the current direction? If it’s on hold, is there a version of it you’d still want to build if the resources were there?

I’ve shut it down completely. As I said, the first version was just a UI button. I realised there was no point building something with no real foundation underneath it — something no one could actually use in practice because of smartphone hardware limits. I killed the project because in business, you can’t hold on to dead ideas.

What’s the current state of the Air Canvas / AI workspace product — how many people have actually used it, and what feedback has surprised you most?

I’ll say the least popular thing a startup founder can say: right now I have zero real active users. The technology is fully built, the code is written, gesture control through the webcam works well. But in the real world, if you don’t have money for advertising and marketing, nobody sees your product.

I’ve built a great engine, but I don’t have fuel — a traffic budget — to make it move. That’s the harsh reality of being a solo founder.

Part 5 — What’s next

What does the Hero Training accelerator application represent for you right now — what would change if you got in?

I applied — and I’ve already been turned down. And that rejection proves, once again, exactly what we’ve been talking about: they want people with money, with visas, with IELTS certificates. European accelerators look for safe, standard cases. I don’t fit their template right now.

If you could get one thing tomorrow — funding, a mentor, a travel grant, a partnership — what would move things forward the most, concretely?

“I don’t need abstract advice or mentorship. If I could get one thing tomorrow, it would be a micro-grant — seed funding — specifically for marketing and server capacity. I can write any code and build any interface myself. I just need a starting budget to put Air Canvas in front of the first thousand real people and get actual metrics back.”

What would you tell another 18-year-old in a small town somewhere who has an idea but assumes they need to be in a city to build it?

I’m not going to give him Silicon Valley motivational fairy tales about ‘believe in yourself and it’ll all work out.’ I’ll tell him the truth: where I’m from, you have three standard doors.

The first is finding a stable, boring job as a manager. The second is going to college or university for a diploma. The third is going to a construction site to carry bricks — which is genuinely how a lot of people around here survive.

A startup is the secret fourth door, and it’s the cruellest one. You’ll have nothing. No proof it’ll work, no one using your product at the start, and nobody will care. Only go into startups if you literally cannot live without writing code. If you can live without it — go carry bricks instead. At least that pays reliably.

---

Solifon AI is available at solifonai.online. Follow StartupReporter.eu for more on what access to Europe’s startup ecosystem actually costs the founders outside it.

This article was drafted with AI assistance based on a live interview conducted by startupreporter.eu, then edited by the startupreporter.eu team.