I build agents that do real work, with a person approving what they change. AutoAgency runs a team's agents on a schedule, logs every run and holds changes for review. At Zilliqa I shipped the first AI bots and an MCP server with 3k downloads in week one.
I design and ship with coding agents in the loop, from Figma to production in Next.js and Supabase. Red Clay, LTIN and HikmaAI went live this way. I also publish open-source Claude Code plugins so other teams can work the same way.
I help teams decide where AI earns its place: which workflows to hand to agents, where people stay in control, and what to build first. For an early-stage AI startup I turned the research into one thesis the founders could all tell.
