Product engineer — agent infrastructure. London.
I build end-to-end systems for AI agents: data pipelines, MCP servers, and the plumbing that gets current, structured context into a model at inference time. iOS background; now focused on agent tooling.
A curated knowledge base your agent queries at inference time. 40 sources scraped, AI-summarised, and structured into agent-ready JSON — ~15× less context than raw sources for the same signal. Ships as an MCP server for Claude Desktop, Cursor, and Claude Code.
Stack: Python · GitHub Actions · DeepSeek Flash · Supabase (pgvector) · Cloudflare Pages Runs fully automated on a Mon/Wed/Fri cron for ~£2–3/month.
Pick a project type, answer a few questions, get a monthly price across providers (Anthropic, OpenAI, Google, DeepSeek). No tokens, no jargon. Python + FastAPI, CI-run test suite.
→ Repo · Try it live
Facts with validity windows, per-agent access control, provenance-gated writes. Runs daily as shared memory for multiple agents. Open-source release in progress.
Python · Swift / iOS · MCP · Supabase / Postgres · GitHub Actions · LLM pipelines