In this episode of the Sunny Ray Show, host Sunny Ray talks with Matas, a Brazilian founder whose career began in mechanical and oil and gas engineering before a university robotics thesis pulled him toward software. After freelancing, he joined a crypto company in Sao Paulo, rising from junior developer to CTO within four years, an experience that led him to found the venture studio Venture Miner in December 2019. Since then his team has launched seven ventures spanning NFTs, carbon credit tokenization, fintech for pregnant women, AI sales tools, and more, learning hard lessons about distribution, regulation, and grit along the way. The conversation then shifts to artificial intelligence, tracing his path from using neural networks for engineering simulations in 2016 to witnessing the rise of transformer based language models. He discusses how AI has reshaped how he teaches coding, embracing vibe coding while still emphasizing sound software architecture, and previews Clutch, his new AI powered distribution product. The episode closes with plans for a follow up conversation covering robotics, blockchain, and AI in more depth.
An oil and gas engineer turned crypto CTO and AI founder shares his winding path from robotics to Web3 to vibe coding.
Before discovering Bitcoin and AI, what was your background?
I started as an engineer in Brazil, in oil and gas, a major industry where I live. My graduation thesis was in robotics, building automated equipment for oil field cleaning and navigation. That got me hooked on the software side more than the mechanical side, so I kept studying software through a master's degree in computer simulations, modeling thermal and fluid flows by writing code.
What got you interested in robotics, and how long ago was this?
It came from a university program where we used Lego kits to teach programming, originally aimed at getting kids into coding. I joined as a side gig, not really pursuing robotics on purpose, but I discovered the field was huge. What hooked me was the short development cycle, you could fix a robot's behavior in minutes instead of the months traditional engineering projects took.
How did you first get into crypto?
After graduating, engineering jobs were scarce in Brazil because industry was declining. I started freelancing and landed a gig at a crypto company in Sao Paulo trying to build a commodities exchange between Brazil and Africa using crypto payments, similar to stablecoin rails today. I went from junior developer to CTO in under four years, then left in 2019 to start my own company.
What is Venture Miner, and where does the name come from?
We mine ventures, like miners collect transactions into a block, we collect ideas and talent and turn them into companies. We teach people to code, not just to get hired but to build their own ventures. I started it in December 2019, right before the pandemic, after running a Web3 agency and boot camps for a few years first.
What ventures has Venture Miner built over the years?
We have done seven so far. An AI to write books and an email automation generator with our friend Charles, ShopMaven which automates WhatsApp sales, NF Boost during the NFT boom, a horse racing platform called Pistister, ESB for pregnant women, and TakuCarbon, a tokenized carbon credit marketplace. None has been the huge breakout yet, but the journey has taught us a lot.
What's the biggest lesson from all these ventures?
It is all trial and error. You think you know how things will go, then reality hits differently. Distribution is hard, selling is hard, regulations get in the way, founders break up, employees leave. The common factor in every successful startup is grit, the strength to endure and build something lasting despite the world tearing your plans down.
How did you first encounter AI, and how has your understanding evolved?
In 2016 during my engineering studies, I used neural networks as a statistical shortcut for solving complex fluid flow equations with too many variables. When GPT2 and later GPT3.5 emerged applying similar deep learning to language, I was surprised it worked at all. Now models like Claude and Gemini feel near human level, even though technically they are still just predicting the next token.
How has AI changed the way you teach people to code?
It is about ninety percent of what we do now. People are not learning to code anymore, they use AI as a shortcut, and honestly that is a losing battle to fight. So instead we teach architecture and best practices around the code AI generates, helping people avoid leaking client data or breaking payments while still building real ventures fast. That is the vibe coding approach behind our new brand, Vibe Pieces.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interview