sergioavi.la
Index3 systems

Work

Systems I've designed, built, and run.

Headroom

LLM gateway and control plane — virtual keys, atomic budgets, provider failover, per-request cost. And the measurement almost nobody publishes: at the default threshold, a semantic cache answered 98 of 130 unseen questions from a neighbour — 92 of them wrong.

01 · Architect · Sole Engineer · 2026Active
Python 3.12FastAPIasyncioPostgreSQL 16pgvectorDynamoDBNext.js 15TypeScript
Backline

Agent platform for music label operations — 3 agents over 468K royalty lines, a 133-question exact-answer eval suite gating CI, and a measured model ladder: the flagship tied the mid-tier at 4× the price.

02 · Architect · Sole Engineer · 2026Active
Python 3.12FastAPIPostgreSQL 16pgvectorTypeScriptvLLMDocker ComposeGitHub Actions
Prometheus

Self-hosted autonomous entity — drives, persistent memory, a dreaming background mind — instrumented on real hardware and debugged at the behavioral level. 1,296 tests.

03 · Architect · Sole Engineer · 2026Active
Python 3.11FastAPIvLLMChromaDBReactWebSocketsPytestGitHub Actions