AI Engineer · Freelance
08/2025 — Present
- Designed and built Umbral, an iOS/Android mobile app with an LLM-based interpretation engine, on a hexagonal architecture.
- Implemented RAG with semantic search and document memory over Supabase and pgvector: contextual chat and symbolic analysis.
- Built an AI quality evaluation gate using embedding similarity that breaks the build when quality drops below threshold.
- Added AI security: guardrails, fail-closed moderation, privacy by default and anonymization.
- Automated Google Ads campaigns with Node.js and OpenAI for bid optimization and reporting, integrated into an existing codebase.
TypeScript · Node.js · Fastify · OpenAI · Supabase · pgvector · Qdrant · React Native · Docker