Arti Assistant
Goal: Build an AI-powered educational agent that helps educators query a curated knowledge base on consent, digital safety, and sexual violence — without hallucinating sources.
Challenge: Implementing a full RAG pipeline from scratch — chunking 14 documents, generating embeddings, storing in pgvector, and performing semantic vector search via Supabase Edge Functions — without a pre-built framework.
Outcome: A working chat interface that retrieves contextually relevant content using semantic similarity, with source attribution and a pnpm monorepo architecture.