I build production systems at the intersection of AI automation and quantitative finance: the kind that run unattended, handle real users or real market data, and fail safely.
Most of what I build is production or client code, so a fair amount lives in private repos. Here's the work.
💬 Remio, a B2B WhatsApp AI receptionist · built, deployed, currently paused An AI receptionist for cleaning businesses that handles bookings, quoting and staff assignment over WhatsApp. FastAPI and Supabase backend, Llama 3.3 for the language layer, deployed on Railway, with a Next.js operator dashboard on Vercel. → Shipped to demo-ready and multi-tenant, then paused. Code private (commercial).
📊 tradingbotmimi, an algorithmic trading bot · public Python bot that paper-trades US equities on live Alpaca data using a multi-signal engine (MACD, RSI, volume) with real risk controls: daily loss cap, kill switch, broker reconciliation, and a watchdog for unattended reliability. Run with production discipline, including walk-forward backtests and a costed go-live gate before any real capital.
🧾 Financial reconciliation tool · client work Built during an internship at ComAfrique to automate a manual reconciliation process, delivered and in daily use by the team. → Built for a live client. Code private for confidentiality.
More projects and my finance work are on my site → meet-thaker.vercel.app
Python · FastAPI · TypeScript / Next.js · Supabase / Postgres · LLM app development (Llama 3.3, Claude) · Railway · Vercel