I build entire AI-infrastructure systems solo — from a pure-Rust LLM inference engine up to a self-hosted DevOps platform that AI agents run end-to-end. One rider, a whole posse of models and agents doing the legwork. Not a demo — I've run the gang in production across 4 commercial projects, and that's where Kovanex was born.
Field notes on solo development in the AI era — real commercial builds, real numbers, and the methods that keep one engineer and an AI gang on course. New dispatches on the way.
A key proves who is speaking; nothing today proves on whose authority. The delegation layer for working agents — the principal signs, a witness countersigns, and chains end because relationships end. Your SSO provider is already the witness.
MeasuredNo API keys, no integrations — the agent operates the running app itself, and keeps what it learned: a dictionary of the app's controls and replayable operation paths. First Telegram run 110 s, second 59.9 s — and a consent gate no skill can learn around.
FeaturedEveryone became a solo founder by renting the stack. I built the stack — server, database, Rust ML training — and manage the one AI agent that runs it.
AI now lets one engineer deliver what used to need a team. I ride the full range — from low-level ML kernels to platform architecture to the leadership call — with an AI posse doing the grunt work under my orders.
Pure-Rust LLM inference engine with Apple Metal GPU + quantized edge kernels, an embeddable HNSW vector database for RAG, and autonomous agents — brain, control plane, execution runner.
Rust · LLMOps · RAGKovanex — a self-hosted DevOps platform in Go: Git, CI/CD, Kanban, registry, a Shamir secrets vault, and an MCP server giving agents end-to-end control. Built solo, dogfooding itself.
Go · gRPC · self-hostedIndependent SRE for Web3 & fintech: cloud exit at −85% cost, a hot query cut 2,500ms → 0.8ms (~3,000×), HA PostgreSQL/Patroni + Vault load-tested to 11,000+ req/s.
SRE · Postgres · VaultI ran an enterprise IT operation of ~90 people and a ~1 billion ₽/year budget (≈ $30M/yr at 2013 rates), then deliberately went hands-on for the Web3/AI frontier. MIPT MSc, Engineer-Physicist. Russian (native) · English (full professional) · Serbian (working).