Aminjon Azizov

Backend & Full-stack Developer · Python, Go · AI-assisted development

I design, build and run production systems: backend services, Telegram bots, an Android client and the Linux server fleet behind them. AI coding agents (Claude Code, Codex) are my main tools — I own the architecture and the task breakdown, review and test what the agents write, debug in production and keep the service running.

Tajikistan · open to remote full-time roles · Russian and Tajik (fluent), English (basic)

Commercial experience

SomonVPN — a commercial VPN service, built and operated end to end. Closed source; case study on aminyx.top.

• Backend — Go modular monolith, REST API, PostgreSQL: accounts, subscriptions, payments, server catalogue, client config generation
• Bots — Telegram bots in Python for sales, payments and support
• Android — Kotlin / Jetpack Compose client with WireGuard and Xray tunnels
• Infrastructure — Linux server fleet (Xray VLESS/Reality, WireGuard, QUIC) with automated provisioning, health checks and monitoring
• Aminyx Link — an experimental transport protocol in Rust used inside the product; closed source, not independently audited

Client project — a turnkey VPN service for a client: website, user dashboard, Telegram bot with payments, and the servers.

Selected projects

Project	What it is	Stack
aminyx.top	Personal site and portfolio: three languages, three.js scenes on a single WebGL context, Playwright smoke tests in CI	JavaScript · Vite · three.js
somoni-tracker	Expense tracker: Telegram bot for one-line input in RU/TJ/EN with typo-tolerant categories, plus a web analytics dashboard	TypeScript · Next.js · grammY · Drizzle · SQLite · Docker
Cybersec	Free cybersecurity course from zero to junior in Russian and Tajik, with a quiz site and CTF tasks — site	Python · GitHub Pages
keelstore	Experimental — a learning project built with an AI agent. Embedded LSM-tree key-value store: WAL, skiplist, bloom filter, levelled compaction	Go
hnsw	Experimental — a learning project built with an AI agent. HNSW vector-search index with an in-browser demo — demo	Rust

Stack

How I work

• Architecture first. I decide the structure, data model and boundaries, then split the work into small, checkable tasks for AI agents.
• Verification, not trust. Every change goes through tests, a second agent as reviewer, and checks on real devices and servers before it ships.
• I run what I ship. Logs, monitoring, incident response and production fixes are part of the job.