Backend engineer. I ship production systems.

I build production-ready platforms with business-oriented architectures – from websites and backend services to microservices and LLM-powered solutions. I also ship desktop software for Windows, Linux, and macOS. Open to full-time, contract, and project work worldwide.

artemskir.com · X · LinkedIn · Telegram · artemsskir@gmail.com

• Starting a Master’s and taking the university journal platform further – the backend is already used by web and mobile clients in the university test contour.
• Operating SRbots – a cross-platform desktop product I’ve owned since 2022.
• Actively looking. Backend / platform roles. Remote, on-site, or relocate – anywhere.

Electronic journal: attendance, grades, assessment forms, async PDF/Excel reports.

Two independently deployed services behind Nginx. Edge validates Keycloak JWTs, enforces roles, and routes. Domain owns business logic, PostgreSQL, audit, and background workers. Edge never touches the database. Domain never authenticates.

• 94 API routes, 4 roles, OpenAPI-first (oapi-codegen) and type-safe SQL (sqlc)
• Load-tested: 500k+ requests, ~1000 req/s, avg ~5 ms, p95 ~13 ms (4 Edge replicas, gzip)
• Observability: Prometheus, Grafana, Loki, Tempo · CI/CD deploys on push to master

Go Gin PostgreSQL Keycloak Nginx Docker Prometheus

Shipped and operated since 2022. Native GUI on Windows, Linux, and macOS. Docs: how it works.

Concurrent Minecraft protocol clients for 1.8–1.21, including BungeeCord / Velocity. I own architecture, releases, and distribution.

• Connection lifecycle: register/auth, auto-reconnect, proxy lists with health checks
• Pathfinding (smart / fast / standard) – fast mode cuts route planning ~50×; standard uses ~250% less RAM
• PVP/PVE, chat and command automation, logging, config import
• Scale is hardware-bound: bot count, render distance, and proxy uniqueness are first-class settings

Desktop Networking Concurrency Proxies Wails Windows Linux macOS

Six-class semantic segmentation on aerial imagery (roads, buildings, vegetation, trees, cars, clutter).

U-Net + ResNet34 (ImageNet). FastAPI inference + Streamlit UI.

• Val: IoU 71.56%, F1 82.73%, accuracy 87.22% (buildings 97%, cars 99%)
• ~250–450 ms / image on Tesla T4 · ~180 rpm with 4 workers

Python TensorFlow FastAPI U-Net

Backend: Go · Python · Java · PostgreSQL · Redis · Gin · Spring · Django
Infra: Docker · Nginx · GitHub Actions · Prometheus · Grafana · Loki
Desktop / web: Wails · React · Node.js
ML: TensorFlow · FastAPI

• 2022–2026 – B.S. Information Systems and Technologies
• 2025–2026 – Geoinformatics Development with AI Methods
• 2026 – Mathematical Modelling and Applications (UCI, Cuba)
• 2022–present – SRbots (owner)
• VK IT-Diving finalist · Moscow Mayor’s Contest · DevSecOps 2024

Russian (native) · English (B2)

artemsskir@gmail.com · LinkedIn · Telegram · X
DM me. I read everything.