Description
We're building an MVP for a SaaS product where an AI agent analyzes legal documents (rental agreements / insurance policies — India market) and flags risky or unfair clauses in plain language for individual consumers. We need a full-stack developer with genuine hands-on experience building LLM-powered applications — not just general web development. This is the core of the project: a retrieval-augmented pipeline (OCR → clause segmentation → embedding-based retrieval against a structured risk-pattern knowledge base → LLM analysis → structured JSON risk report). **Tech stack we're planning around (open to your input):** Next.js (frontend), Python/FastAPI (backend), PostgreSQL + pgvector (database/retrieval), Anthropic or OpenAI API (LLM), managed OCR API, Razorpay (payments), Clerk/Supabase Auth, Vercel + Railway/Render (hosting). **You should have:** - Shipped a real product (not just a personal project/demo) using an LLM API with structured/JSON output - Experience implementing embeddings + vector similarity search (pgvector or similar) in production - Comfort with Python for backend/pipeline work and React/Next.js for frontend - Bonus: experience with OCR/document-processing pipelines on real-world scanned documents - Bonus: any experience in legal-tech, fintech, or other trust-sensitive consumer products **Scope:** Full MVP build as described — document upload/OCR, retrieval pipeline, LLM analysis with confidence scoring, report UI, auth, payments/subscriptions, basic admin