Built an end-to-end AI-powered vehicle intelligence system that ingests real marketplace data (Otomoto, OLX), normalizes listings, applies hybrid deterministic + AI scoring, and surfaces ranked, risk-aware insights through a searchable dashboard.
Role: Product Architect • Backend Engineer • AI Systems Engineer
Key Capabilities: • Real data ingestion via Apify (no mock data) • Unified listing schema across marketplaces • Advanced search & filters (make, model, price, mileage, fuel, condition, etc.) • Natural language search → structured filters • Deterministic 0–100 scoring for fast, cheap ranking • Selective AI analysis (text + image) for defect, risk, and anomaly detection • Authentication, multilingual support (EN/PL), CSV export
Architecture Highlights: • Modular, extensible MVP designed for scale • AI used only where it adds signal (cost-aware design) • Supabase backend with Edge Functions & RLS • Adapter-based ingestion to handle API constraints and swaps
Roadmap: • Voice AI agents (outbound & inbound) • CRM handoff and booking automation • Queue-based orchestration
One-line summary: Turned noisy vehicle marketplace listings into actionable intelligence using real data, hybrid AI scoring, and a scalable SaaS architecture.