Daya

An AI scam detection platform built for the United Nations Cybercrime Hackathon finals in Hanoi, Vietnam. Daya was selected as one of 8 finalist teams out of 50 across Southeast Asia and classifies emerging global scam reports using government and crowdsourced intelligence.

Roles

AI Engineer

• Engineered a full-stack AI platform using TypeScript and Node.js to classify emerging scam reports
• Developed an operational RAG pipeline using OpenAI embeddings
• Queried aggregated government and crowdsourced reports stored with metadata inside vector-optimized Supabase tables
• Normalized volatile incoming text segments into unified JSON schemas

Technical Highlights

• Built vector-optimized Supabase tables for reports, metadata, and embeddings
• Applied structured chunking, micro-validations, and LLM-driven categorical enrichment
• Used evidence-backed retrieval to support classification and recommendations