Production AI agents
Autonomous agents that act safely when the stakes are real - exactly-once execution, fail-closed actions, and self-improving accuracy loops. The hard 20% most demos skip.
proof Valory · Vacya
Portfolio · 2026
AI Engineer · Tech Lead · ex-Startup CTO
I build AI-native products end-to-end — and ship them fast.
I ship the whole thing — backend, frontend, infra, and the AI layer on top.
AI Engineer and ex-startup CTO — I take products from zero to one and own them end-to-end. I co-founded Quido (0 → 5,000+ users), shipped production autonomous agents at Valory, and now lead engineering at Spocket. I work AI-native: my own toolchain of skills, hooks, and subagents clears the drudgery, so shipping fast is the default — not the exception.
I build agents for a living, so my portfolio has one. It answers the questions a hiring manager actually asks — try one below.
Offline · answers on-device · no data leaves your browser
select an entry to expand
tech lead · jun 2026 - now
full-stack engineer · jun 2025 - 26
Not a service menu — things I've already shipped and can own end-to-end. Each one is backed by real work, not a promise.
Autonomous agents that act safely when the stakes are real - exactly-once execution, fail-closed actions, and self-improving accuracy loops. The hard 20% most demos skip.
proof Valory · Vacya
Whole products from an empty repo to real users - backend, frontend, mobile, infra, and the AI layer on top. Founded, owned, and shipped end-to-end.
proof Quido · Thimblerr
My own agentic dev layer - skills, hooks, and subagents - that clears the drudgery so the whole team ships faster. 35,000 alerts across 100+ repos, cleared in ~2 weeks.
proof Spocket · Valory
An autonomous trading agent on Polymarket (Polygon), led end-to-end: strategy design plus a self-improving accuracy loop that scores predictions against real market outcomes and feeds them back. The hard part was the safety layer - exactly-once execution and fail-closed actions.
An autonomous DeFi agent managing on-chain positions on Optimism - concentrated-liquidity optimization, asset whitelisting, and fail-closed secure withdrawals with a trailing stop-loss exit. Shipped to production; shown at ETHCannes.
A production Telegram AI personal assistant, delivered solo for a client on the Hermes framework - scoped and architected end-to-end, from first requirements through a live client handover.
Users invest in expert-curated strategies by linking their exchange. I built the whole stack - FastAPI + Spring Boot + AWS backend, React Native mobile, Next.js web, automated portfolio rebalancing - and scaled it 0 → 5,000+ users, $50k+ AUM.
A voice AI ordering agent on ElevenLabs that handles a full drive-through order by voice - upsells, allergies and removals, knowledge-base lookups for availability, and human handoff when it should step aside.
My own agentic dev layer on Claude Code - custom skills, hooks, and subagents for code review, security review, dependency cleanup, and migrations. At Spocket it cleared a 35,000-alert security backlog across 100+ repos in ~2 weeks.
Open to contract and founding-engineer roles, remote-first. If you're shipping something that matters, I'd like to hear about it.