Godot

Godot was born from a concrete HR problem: recruiters were drowning in hundreds of CVs and losing hours manually cross-referencing CVs with job descriptions, with results that were hard to justify. The goal was to turn that search work into a conversation.

The platform ingests CVs and job descriptions, turns them into embeddings and answers recruiters' questions in natural language, always citing the source. You can ask, for instance, who has Kubernetes experience in finance and get a reasoned answer, with the supporting candidates.

I led product and experience end to end: from the first POC on Google Cloud, with an ingestion pipeline and a vector RAG engine on ChromaDB and pgvector, to the port onto a multi-tenant Next.js and Supabase architecture with RLS and commercial plans.

The experience is built around three roles — HR, candidate and administrator — with Google SSO, WebSocket response streaming and support for 99+ languages, because a modern recruiting tool must speak the language of the people using it.

On the technical side it was an exercise in orchestration: LangChain for retrieval, FastAPI for services, Terraform for the GCP infrastructure. The result is screening up to ten times faster, where instead of reading you query the archive and get cited answers in seconds.