Dropping out TUM to build & ship. Meet Robert.
Robert dropped out of TUM when ChatGPT came out. He joined a hacker house, went viral for his local LLM demos and launched Fixkey.ai. He just got into YCombinator with his new project, 1nterface.ai
Studies CS Exchange @TUM (drop-out) ‘23, Systems Engineering @ Wroclaw University of Science & Tech ‘23
Companies stormy.ai (formerly 1nterface.ai) (YC S24), Fixkey.ai Experiences Loona, ngrow.ai
Origin Belarus
Links LinkedIn, Twitter
Growing Up in Belarus and Moving to Europe
Where are you originally from and how did you end up in Europe?
I’m originally coming from Belarus. When I was 16, the Polish government sponsored my scholarship and allowed me to move from Belarus to Poland to study for free. They gave me a grant and paid for everything. The Belarusian citizenship is like the worst you can have right now, so I was using university to stay in the European Union and fix this problem.
What were you working on in high school in Belarus?
I was in a very special environment in high school. My classmates were like the best olympiads in math and physics in the whole country - extreme high intellect, the highest I’ve met so far in my life. While my friends focused on front-end because it was easy to earn money, I started doing machine learning courses on Coursera, like Andrew Ng’s stuff. I was 16 and made all those certificates, but no one around me was talking about ML or AI. I was like the first one saying “Hey guys, have you heard about ML? Do you think it’s cool?” and they’d reply “Dude, this is nothing, you can’t earn money with this.” I got a little disappointed in myself, thinking my vision was wrong.
University Experience and First Startup Jobs
Why did you decide to go to university in Poland?
It was mainly to move out of Belarus, the environment there was not one to live in, literally. When I moved to Poland, Covid was at its peak and everything closed up. It was a complete mess, no one knew how to teach remotely. I took this as an opportunity and got my first job at a startup while studying.
Tell me about your experience working at Loona.
The dude who started Loona was from my high school and had previously sold his startup, which made the technology behind Instagram’s masks and effects, to Facebook. I just cold-dm’ed him and pitched myself for an internship, and he loved my energy. One week after I turned 18, I got the job offer.
While studying in Poland, Loona literally paid me to go back to Belarus for 3 months while I was still doing university stuff remotely. It was the coolest experience - I had just turned 18 and was going to a city I’d never been to - for 3 months - and they paid for everything. It was so dope. I worked there for a year doing Python backend engineering and learned a lot of good practices.
What did you work on at Ngrow.ai?
After Loona, an investor introduced me to Ngrow.ai because she thought I was cool. They made me an offer and one month after I joined, they got into YC. I was doing data engineering, learning how to send 250 million push notifications per day. Ngrow.ai specializes in push notifications and learning user patterns to increase conversion and retention. For example, if you prefer to order Uber Eats in the evening but forget one day, it’ll send you a notification to remind you.
Studying Abroad in Germany and Continued Growth
Why did you decide to go on exchange to Germany?
I figured out my university in Poland sucked and thought about what the best thing I could do was if I wanted to continue studying just for legal reasons. I looked at the ratings and TUM was the highest in Europe. Even though Munich is expensive, I committed to it and decided I’d make it work no matter what. It was the best option for me.
How was your experience at TUM and in Munich?
The atmosphere of building excellence at TUM was pretty cool. I definitely learned some stuff and takeaways on how I can work on my own things better. The way they teach is alright, the way they ask you to answer exams is pretty bad though, very painful.
After 3 months on exchange, I figured out university at TUM was also not so good. I found another job at a startup working on hydrogen optimization and usage. But I didn’t like the management so I left after 2 months. ChatGPT had just hit and it was too big to ignore. Even though I really needed money and Munich was taking every penny I had, I decided to leave the company and start working on my own ChatGPT projects. It was still very early days.
Building AI Projects and Going Viral
What kind of AI projects did you start working on?
I started working on a summarization tool for customer development meetings. I did like 15-16 customer interviews and made almost $300. But then I noticed VC-backed companies got millions in investments for summarization tools. I realized as an 18-year-old with only 3 years of real experience, no one would invest in me. So I made the hard choice to stop pursuing it since it would die in the extremely competitive market.
I got really excited about hackathons and thought maybe they could give me new ideas. My team won a hackathon from Google in Berlin and got 4th place at the SuperAGI hackathon. I met great people like my Alex Pokras (Editor note:Robert’s current co-founder) through this, and learned a lot.
You went viral for posting locally run AI models. Tell me more about this.
I made an open-source Chrome extension to chat with any webpage. You put in your OpenAI API key and can trigger a chat interface on the page. It indexes everything in the browser, keeps your data safe, and lets you chat with the page for free. It got 50K views on Reddit and over 200 stars on GitHub.
Then I made my first demo that went extremely viral - it was like FaceTime but with a local language model that transcribes what’s going on in the camera. I showed it running locally and people loved it - Andrej Karpathy from OpenAI and the LlamaGPT founder liked it and it got over 400K views.
Some of my later demos were 100% fake, like one with a yoga app that tells you if you’re doing a pose wrong. No one could check since I had just released it a day ago. I was focusing more on local models to see where it goes.
Joining a Berlin Hacker House
Why did you decide to join a hacker house?
After an AI Tinkerers event in Berlin, I realized the environment around me in Poland wasn’t ambitious enough. Dominic who runs the hacker house said I could join, so I applied, got accepted, and immediately canceled my rent in Poland. Two weeks later I moved to Berlin and have been living there for 2 months.
If you’re surrounded by people who are stupid, you’re going to be stupid. If you’re surrounded by ambitious people who think they’re going to change the world, you’re going to change the world. The environment and people around you really shape you.
Launching Fixkey.ai, a locally-run sentence-corrector for Mac.
How did you come up with the idea for Fixkey?
After the AI event in Berlin, Raphael, Mathis and I started talking about how it would be cool to type really fast. Raphael showed us his custom keyboard setup and was typing complete bullshit but then triggered a shortcut that corrected it using ChatGPT. We realized this could be a product.
Mathis, the chatpdf.com founder, was onboard. We decided to do a 1-month hackathon to build and sell it. None of us knew native macOS or Swift development but we figured it out together. I even crashed at Mathis’ place for a week while we coded it.
What was the launch like and how has growth been?
We released Fixkey on January 8th and made $200-400 the first day. We kept improving it, doing more marketing, adding tracking and analytics. The first batch of users dealt with an onboarding bug that crashed the whole app, but they kept using it and paying us anyway.
Once we fixed that bug and added more features, people naturally started finding us and making posts/reviews without us paying anything for marketing. Now we’re pushing more into a marketing phase to make the incoming revenue bigger and more stable.
In the last 7 days alone, 250 people have downloaded the app, which is pretty good organic growth.
Productivity Hacks and Optimizing for Longevity
How do you stay productive while juggling multiple projects?
This year I made a promise to myself that once I pick something to work on, I’m going to stick with it until some critical factor shows up. Even if I feel uncertain or don’t like it, I’m going to push through and do it.
I’m also very protective of my time. I usually decline event invitations because even though my schedule looks empty, I’m always busy. I subtract sleep, gym, and eating, and the rest of the time I’m coding.
What’s your philosophy on health and productivity?
I’m an extreme optimizer. I do the gym, follow a blueprint diet by Bryan Johnson, and try to make an extremely good sleep schedule to make my life and productivity span as long as possible. I want to make something great up until the moment I’m physically unable to do anything, whether that’s 70, 80, 90, or 100 years old. Sleep, gym, and eating right - optimizing those is key.
Getting into YCombinator with 1nterface.ai
Tell me about 1nterface, what is your vision?
1nterface is a step towards reimagining how you interact with your computer. Thinking back to the first principle. When you open your laptop, what looks different?
What will you expect to see?
Can AI anticipate your actions and proactively act?
The core idea we go after is knowing your context. Understanding what are you up to in your life as a person.
You're working together with Alex Pokras, how did you first meet and why is this a great team-fit?
I met my co-founder, Alex, at a TUM.AI student initiative a few years ago. We immediately clicked and started talking a lot about AI applications and implications. We won a few hackathons. And even met with google cloud CTO after because of it.
At the time we founded 1nterface.ai, we already knew each other for some time, so it was a smooth sailing from that on.
Congrats on getting into Y Combinator S24, what was the interview process like?
We have been in touch with Garry Tan (President of YC) before I even applied or considered applying to YC.
After that he multiple times quoted my tweets and I just reached out on his opinion about what we are up to.
We have been in touch since December / January.
YC was a logical continuation after we started to have more serios intentions and big thought about how we are going to change the world.
Didn’t feel like a jump. More like a logical step forward.
Hey - Arnie here!
Fyi - 1nterface.ai just launched! Go try out Robert’s project and support them in the early stages. They’re building something super cool.
Also - Sorry for the inconsistent posting schedule, I’m hoping to get it under control eventually (there’s really so many more cool interviews that are already recorded and I just need to get through 😅 ).
Just had some super busy few months (I have Twitter where I post my daily activities, go check it out) 🙃
Hope you enjoyed this blog - as always, let me know what you think!
Initially I thought that Robert was chasing the "AI hype", but later into the blog I was pleased to found out that his projects could actually turn into a viable business that has an impact on the regular non-tech user. Good job ⚡️
Alex and Robert have a great future ahead. Let's go guys, TUM.ai support you at every step of the way! 🚀