Study Activity Dashboard

A full-stack “Study Activity Dashboard” that lets a learning officer see how active a study group’s students are — built end-to-end for a software-engineering take-home.

• Built a typed FastAPI backend over PostgreSQL using asyncpg with raw, parameterised SQL organised behind a repository layer, computing all aggregation (per-student stats, a gap-free 14-day activity timeline) in the database via GROUP BY / AVG / COUNT rather than in application code.
• Enforced the core business rule (quiz scores required and constrained to 0–100; null for every other activity type) in two layers — a Postgres CHECK constraint and a matching Pydantic v2 validator — with correct 201 / 404 / 422 status codes.
• Containerised Postgres with Docker Compose, auto-running the schema and a deterministic seed on first boot so the whole stack comes up with a single command; secured the API with a shared X-API-Key header.
• Built the frontend in React + TypeScript + Tailwind (Vite), using TanStack Query for data fetching and cache invalidation, Recharts for the activity chart, and a sortable table with real submit UX (in-flight disabling, toasts, error recovery).
• Added an “At a glance” LLM summary endpoint that computes the numbers in SQL and has Google’s Gemini only reword them, with a graceful templated fallback so the API degrades cleanly instead of failing when the model is unavailable.

Source: private take-home repo