LegalAid ↗
An end-to-end GenAI workspace for advocates
AI Architect · AI Product Manager · AI Engineer
Outskill AI Generalist Fellowship — a six-month programme. Built on weekends with one other fellow.
Advocates lose billable hours to paperwork that arrives handwritten, typewritten, stamped and damaged. LegalAid gives them one workspace for documents, clients, cases, teams and outreach.
- Document processing — An n8n pipeline takes raw scans through OCR, cleanup and classification, and moves each document along a tracked status pipeline rather than leaving it in a folder.
- Retrieval (RAG) on Pinecone — Legal questions are answered against the firm's own documents in real time — embeddings indexed in Pinecone, so the answer cites what the practice actually holds instead of what a model recalls.
- AI avatar assistant — HeyGen and ElevenLabs drive a live avatar that acts as a legal assistant for junior practitioners — the part that turns a document tool into something a new advocate will actually ask questions of.
- Data layer in SQL — Postgres on Supabase in production — schema, row-level security policies, triggers and seed data written in SQL — with a SQLite path so the whole app runs locally with no cloud account.
Spec-driven and test-driven throughout — 104 tests across the front and back ends.