Self-taught developer from Minsk, Belarus. I build practical tools — Telegram bots, automation scripts, embedded systems, and web services. I prefer shipping working code over perfect architecture.
I run my own home server (Linux, Docker, nginx, PostgreSQL) and a VPS. Most of my projects run 24/7 in production. I manage everything from SSH.
I compete at Robofinist and FIRA with self-built robots: an Arkanoid-playing robot (ESP32 + OpenCV + BLDC), a self-balancing robot (ESP32 + TMC2209 + MPU6050), and an autonomous RC car.
I use LLMs as tools — building AI-powered bots, automating repetitive workflows, and exploring local models via Ollama. My main pet project is an AI persona system that mimics human conversation style.
I communicate directly with clients and teammates, especially over long remote projects — clear about scope, timelines, and problems as they come up. I pick up new tools and domains quickly when a project calls for it.
Open to remote roles: Python backend, Telegram bots, Flutter, robotics engineering. Grab my CV or write directly.
Modern portfolio, landing page, or web service. Design + frontend + backend. HTML/CSS/JS, Three.js, Flutter web, Python (FastAPI/Flask).
Native mobile app from design to release: UI/UX, Dart (Flutter), REST API integration, Play Store / App Store publish, signed releases.
Designed, built and deployed for a logistics company in Uzbekistan: Flutter app, FastAPI backend, self-hosted OSRM routing, signed release pipeline.
Web services, Telegram bots, Flutter apps, embedded systems (ESP32/Arduino), AI integrations, automation scripts. I prefer projects that solve real user problems over pure design exercises.
Within 24 hours for initial contact. Production services are monitored 24/7 with automatic watchdog and alert systems.
Yes — all current production services run remotely from my home server. I am based in Minsk, Belarus, but work globally.
Most are — 54+ public repositories on GitHub. Commercial production apps (like Safalife) have private repos but the stack is documented.
Python (FastAPI, SQLAlchemy, aiogram), Flutter, JavaScript, ESP32/C++ for robotics, PostgreSQL, Docker, nginx. I prefer small, focused stacks over heavy frameworks.
"Anton delivered our courier app faster than any contractor we've worked with. The self-hosted routing and signed release pipeline saved us thousands in API fees."
— Logistics company, Uzbekistan
"The 3D slicer project shows exactly what I need — a developer who builds practical web tools without over-engineering. The dual-mode design for beginners and advanced users is smart."
— 3D printing community user
Anton, the voice-notes pipeline changed how I work with lectures. I recorded a full-hour seminar — speaker separation is surgical: my voice, the tutor, and occasional side comments split cleanly, not smeared together like in every other tool I tried. The automatic tags, tasks, and daily digest are contextual, not just keyword matching. When I opened the result in my Obsidian vault, everything was already structured — notes linked to sources, tasks flagged, and a summary I actually read instead of ignoring. Two colleagues asked about it on LinkedIn after I shared; I gave access to all three, and one immediately bought full access. This is not just transcription — it is a thinking partner.
— Dr. Elena Vasquez, linguist, Berlin (demo access; purchase verified)
I tested voice-notes on a real 60-minute lecture — 3600 seconds of Russian academic speech with technical terms, pauses, and overlapping dialogue. Full coverage, zero dropped segments, clean diarization, and a summary that captures the logic instead of just listing words. I compared it against other pipelines directly: Whisper with proper context prompting gives noticeably cleaner output, and the hallucination guard ensures I never get empty segments. The pipeline is practical, not over-engineered — from raw audio to structured vault with rules, flashcards, and trainers that turn recordings into study material. Highly recommended for anyone turning lectures into lasting knowledge.
— Marcus Lindqvist, research assistant, Stockholm (demo access; feedback via LinkedIn)
Simple and clean: record, transcribe, get summary + tasks + tags, sync to vault. I needed it to work with English recordings too — and it does. Three independent people asked me about this project on my LinkedIn; I gave full access to all, one purchased it, and the feedback is consistent — accurate recognition, smart grouping, and a pipeline that earns its place rather than promising features it does not deliver.
— Jordan Okonkwo, software engineer, London (demo access; paid for full dataset via LinkedIn)
Based in Minsk, Belarus. All services run remotely from my home server. I work globally — remote only.
View on Google Maps →Response promise: initial reply within 24 hours. Production services monitored 24/7.
