Odigos

Overview

Odigos is a retrieval-augmented chatbot for families in Italy living with autism, built for Fondazione Paideia during my time at RADIAITE. It answers from retrieved sources rather than from whatever a general model remembers, and it runs in production.

What I built

I built the pipeline end to end: knowledge-graph traversal over Neo4j, vector search over pgvector, semantic reranking with Cohere, and answer generation with GPT-4.1. The product grew to more than 58 features.

How it works

A question takes two retrieval paths: a traversal of the knowledge graph and a vector search. Cohere reranks the combined candidates so the model sees a short, relevant context, and GPT-4.1 writes the answer from it.

Deployment

The service runs on Azure Container Apps with Neo4j and pgvector behind it, and every change ships through a CI/CD pipeline on GitHub Actions.