I like understanding things.
- Sometimes by building them.
- Sometimes by taking them apart.
- Sometimes by going somewhere I've never been.
IoT student in Plzeň. AI engineer at Medicalc, co-founder and CTO of nerve labs.
Nothing interesting happens in a perfectly smooth universe.
For 380,000 years the universe was fog. When it cleared, it left this: the oldest light there is, warmer or colder by one part in 100,000. Every galaxy grew from a slightly warmer spot.
Then nothing shone for the next hundred million years or so.
Gravity worked anyway.
The scaffolding nobody sees.
Dark matter holds galaxies together without giving off any light. Mine is books, papers and datasheets at 2 AM. And the other half of learning: 3D printing something that probably didn't need to be 3D printed, soldering something together just to see if it works.
The constants this universe runs on
- Understand the system before optimising it.
- Prefer simple architectures over unnecessary complexity.
- Build things that can actually be used.
- Automate repetitive work.
- Document what others need to reproduce.
- Open-source whenever possible.
- Learn by making.
- languages
- Czech, English, and slowly Japanese and Ukrainian
- motto
- make → break → learn → repeat
Every atom of copper on these boards was forged in stars that died before the Sun was born.
I just route it.
From the copper up.
I like going all the way down the stack: a trace on a PCB, the firmware that wakes it up, the protocol it speaks, the app that talks back. My favourite projects are the ones where you can't tell where the hardware stops and the software starts.
- languages
- C, C++, Rust, MicroPython, Python
- silicon
- ESP32, STM32, AVR, Raspberry Pi
- making
- KiCad, Fusion 360, Blender, 3D printing
For SMART-BLINDS I designed two boards of my own: a router that turns Wi-Fi packets into Zigbee, and the end device that listens. SMART-BLINDS →
A drone that should one day fly itself.
An F405 flight controller does the fast, boring, vital work of staying in the air. A Raspberry Pi 5 rides on top as the companion computer. It talks MAVLink, reads GPS and a camera, and decides where to go next.
Before the air there was water: VIRIDIS, a submarine two classmates and I built from scratch. I wrote the software that steers it and listens to it. VIRIDIS →
Three rotors. Two of them tilt.
My newest experiment is a 3D-printed VTOL. The plan: take off like a helicopter, tilt the two front rotors forward and fly on the wing, with a third rotor at the tail holding the balance. It isn't finished, which is the fun part.
What it should do, and where it is now: VTOL →
From six billion kilometres away, Voyager 1 saw Earth as less than a pixel.
I go and look.
A new place does to my head what a fresh pair of eyes does to a stubborn bug. Different trains, different sockets, different answers to the same questions, and usually some interesting technology on the way.
Poland is Angloville: week-long English immersion camps where I mentor, one conversation at a time. Turkey and China came with competitions.
Plzeňhome, undeveloped
PolandAngloville, undeveloped
Romaniaundeveloped
TurkeyMEB Robot, twice
ChinaWorldSkills, Shanghai
The stars of a constellation are rarely anywhere near each other. People drew the lines to tell stories.
Against the clock.
Competitions and hackathons: a deadline, a room full of people who are better than you at something, and a thing that has to work by Sunday.
Competitions
WorldSkills Shanghai 2026
The 48th WorldSkills Competition, the world championship of vocational skills: about 1,400 competitors from 73 countries and regions, in 64 skills.
International MEB Robot Competition
Turkey's big school robotics competition, run by the Ministry of National Education with TÜBİTAK and TİKA. The first time we brought an FPV drone we built ourselves. The second time, in Antalya in May 2026, VIRIDIS ran the underwater course well until vibration shook a power connection loose.
VIRIDIS →
Hackathons
Rakathon
The first Czech hackathon on oncology: 48 hours, 29 teams, at the Second Faculty of Medicine of Charles University.
Hack jak Brno
A national healthcare hackathon. Our team built an operating-room management system that speaks FHIR.
MedicHub →AI Discovery Day
Organised
CAMPFIRE Hackathon Pilsen
My own hackathon. I organised the whole thing myself.
CAMPFIREPlzeň, undeveloped
Pull back far enough and the galaxies line up along threads, with almost nothing in between.
Everything connects.
Fourteen projects and the threads between them. Point at a project to see what it's made of, or pick a technology and follow it back.
- 01VIRIDISunderwater robotics · control softwareC# / .NET, Telemetry & protocols, 3D printing & CAD
- 02Autonomous droneflight controller + companion computerTelemetry & protocols, Microcontrollers, Raspberry Pi, Computer vision
- 03Tilt-rotor VTOL3D-printed airframe, three rotorsMicrocontrollers, 3D printing & CAD
- 04SMART-BLINDSZigbee router · two custom PCBs · web appTelemetry & protocols, Microcontrollers, PCB design, React / Next.js
- 05Engineer Chat Botfine-tuned LLM on a Raspberry PiRaspberry Pi, MCP, LLMs, RAG
- 06KOREXSaaS · conversations into tasksLLMs, RAG, NLP · KG-LLM
- 07Chat_bot_RAGmy first retriever, written by handLLMs, RAG, Python back-ends, Docker
- 08Life Simulationa procedural world run by an LLMLLMs, React / Next.js
- 09MedicHubhackathon · operating-room managementPython back-ends, Docker, React / Next.js, FHIR
- 10EcoQuestsocial game · on-device detectionReact / Next.js, Firebase, Flutter, Computer vision
- 11Productivity apppay for screen time in push-upsFlutter, Computer vision
- 12Screen translatorOCR over whatever is on screenComputer vision
- 13Math test appFlutter · generated exercisesFlutter
- 14Streaming servicea spare Raspberry Pi, for friendsRaspberry Pi
14 projects, 17 technologies, 39 connections
The cosmic web and the network of neurons in your head have measurably similar structure, 27 orders of magnitude apart.
Vazza & Feletti, 2020
So far you've been scrolling through one thought.
AI you can open up.
Not the magic kind. Attention weights, embeddings, a retriever that finds the right paragraph, a neuron that fires only when it has to.
Touch a neuron.
- LLMs
- I build with them, not around them: an assistant fine-tuned for electrical engineering, NPCs that plan their own days, tools wired in through MCP.
- RAG
- My first retriever was hand-written: PDFs and web pages, vectorised, matched by cosine similarity. The latest one reads a whole school's study materials.
- Transformers
- Attention is the idea I keep coming back to: every token deciding what is worth looking at.
- SNN
- Spiking neural networks put time back into computation, which is closer to how neurons actually talk. Currently learning, slowly and with great enthusiasm.
- Computer vision
- A YOLO model on a phone, drawing boxes around litter so a game can trust you. Push-ups counted by a model. Text read straight off a screen.
- KG-LLM
- Knowledge graphs give a language model something solid to reason over. KOREX uses them to turn conversations into tasks.
- MCP
- Giving models real tools instead of more adjectives. One of mine drafts laboratory protocols.
In 1974, Arecibo sent 1,679 bits towards the star cluster M13. They should arrive in about 25,000 years. Email is faster.
Say hello.
If you're building something where hardware, software and intelligence meet, or you just know a good book, write to me.
- email2sabin8.kluk@gmail.com
- githubgithub.com/Plokar
- linkedinSebastián Bořík
Trajectory so far
- Medicalc, AI engineer
- nerve labs, co-founder, CTO
- Bridge Academy, student mentor
- Freelance, full-stack for a private client
- Biorobotec Bionic Robot Teknolojileri, full-stack developer
- NVIAS, methodologies
- VOŠ & SPŠE Plzeň, studying IoT
And then it slowly lets go.
Nobody knows how a universe ends. In Roger Penrose's version it never collapses. It just keeps going, until the cold, empty far future looks exactly like the hot start of the next one.
Thoughts end like that too. They thin out into other people's heads and start something there.
Keep going and you're back on a blank page.