We have a working vehicle listings analytics platform that needs a production-grade engineer to finish it properly. The architecture is in place — what's missing is determinism, reliability, and real implementations where stubs and placeholders currently exist. What you'll be fixing / building: • Rip out all placeholder and random logic — pricing benchmarks and similarity ranking must be fully deterministic (tiered comparable selection, outlier filtering, weighted scoring, tie-breaking) • WebSocket notification system needs to be auth-aware, per-user, and safe to run across multiple instances (async Redis pub/sub or streams) • AI smart review endpoint: replace the mock with real web evidence retrieval → LLM synthesis → fingerprint-keyed cache with TTL + version invalidation • SQL bugs, missing constraints, missing indexes — find them, fix them, migrate cleanly • Async test coverage (pytest + pytest-asyncio), fully runnable in CI • Facebook integration module needs wiring into the core system Stack: FastAPI · SQLAlchemy · Redis (async) · WebSockets · PostgreSQL · LLM API · Pytest This is milestone-based. No fixed hourly retainer — we scope per workstream. When applying, please send: 1. Total hour estimate 2. Per-workstream breakdown (WebSockets / pricing / similarity / AI review / DB / tests) 3. Risks or unknowns you spot from the brief 4. Your proposed milestone structure 5. Testing approach 6. What access you'll need to get started A technical brief with full scope detail is attached — please read it before applying.