◯about me
Yahor Mesherykou
Python Backend / MLOps engineer
Below are two parts: me as an engineer and me as a human. Both are real.
Me as an engineer (sense1)
Python backend engineer with production ML depth: async microservices on FastAPI/Litestar, Kafka, Kubernetes — for real-time and resource-constrained environments. Built the backend for a fleet of 300+ Jetson devices and an 8-machine cluster; I run the full MLOps cycle — from model registry to production inference (~0.6 TB of artifacts per month). Sped up backtesting 280x by rewriting hot paths in Rust via PyO3. A DDD and Ports & Adapters adept, author of dddguard and snitchbot.
Real-time analytics platform
Core of the platform backend: training-data preparation, backtesting, Dash dashboards for analytics and clustering, the full retrain -> rollout pipeline. Backtesting ×280 on Rust/PyO3; a model-and-prediction store with versioning and lineage — any production prediction traces back to its model. A log reader on Python sub-interpreters, Playwright sessions through anti-bot, an 8-machine cluster with CI/CD.
Computer-vision platform
Microservices for 300+ Jetson edge devices under a strict RAM budget. End-to-end inference pipeline: WebRTC ingest (aiortc), frame preprocessing, model serving across the fleet, event delivery, latency and quality monitoring. A fleet manager on the Python Kubernetes SDK: model and binary rollout, reconciliation, health and audit APIs.
BSU · Faculty of Mechanics and Mathematics
Pure mathematics. Thesis: information entropy and Markov chains. Coursework: probability and statistics, ML, SQL, C++, Matlab.
Me as a human (tapo4ek)
Outside of code I am much the same: long distance, drive and a love of systems — just different objects.
Math -> architecture
A big adept of mathematics: I was deep into it at university — the strictest, most abstract kind, and its beauty kept surprising me. But most of that knowledge is honestly hard to apply in practice — so I moved to programming: fields of every flavor, and above all very practical. And of course you still need math here :)
Everything new
I learn new things greedily and widely — and love sharing them just as much: telling, showing, pulling people in. I have a soft spot for people who are true experts at their thing, whatever it is. The best rest from one hard topic is another hard topic.
Unusual solutions
A special thrill is finding a solution that looks unexpected and then obvious: rewriting pandas in Rust, putting an agent in charge of my notes, replacing Sentry with a Telegram bot.
Persistence as a method
When I take something on, I don't drop it — even long after it stopped being fun. Long projects, habits, sport all rest on that. It's not a burst of willpower, it's just that I keep going — until it's done.
Tasty & healthy
I love eating tasty and healthy — and love cooking myself just as much. Slightly obsessed with healthy food.
not just Grandmaster — hit 5500 GM playing Random. Invented the most obnoxious Terran style out there — battlemines
the best and truly team-based RTS: brawls without farming, pure macro play. A shame so few play it and updates are scarce
strictly for company — friends drag me in
Noita, Tainted Tail, Skyrim, the Total War series — building a system that runs without me sounds familiar
Years in the gym: it trains discipline and clears the head nicely. But once I got stronger I realized iron still leaves you wooden — so I went into boxing.
Boxing is a great sport: it builds almost the opposite of the gym — speed, reaction, coordination, stamina. Drilling moves and combos, working in pairs and sparring is genuinely fun. After a spar the head empties completely and I can switch off — but then it starts to ache)) so I've eased off for now and am looking for a new sport.
I've also started running — mornings or evenings. Thinking about jiu-jitsu or another martial art. On the lookout for a new sport.