LawCoach

AI study platform for law students, with five-model essay grading and a timed bar simulator.

Private beta

Problem

Law students get plenty of practice questions and almost no real feedback. Model answers are generic, graded feedback is expensive and slow, and nothing tracks what a student keeps getting wrong across subjects.

Role

Sole designer and engineer, end to end: product, interface, data model, AI grading pipeline, payments, and deployment.

What I built

Built the full-stack platform solo with Claude Code, covering 20 subjects and a full two-day timed bar simulator.
Built a grading council that scores essays across five frontier models in parallel and reconciles disagreements through issue clustering and median scoring.
Built a cross-provider gateway with ordered model fallback, circuit breakers and schema-validated structured output with retry.
Added per-user document retrieval using pgvector so students can study with their own uploaded materials.

Outcome

In private beta.
Built and maintained solo from 2025 to the present.

Stack

Next.js 16 · TypeScript · Tailwind CSS · Supabase (Postgres + Auth) · Vercel AI SDK via Vercel AI Gateway · Stripe (checkout, portal, webhooks) · Vercel