Filippo Piombino

At AlphaProcessControl, I engineered hybrid RAG and GraphRAG systems that made technical plant information searchable in seconds rather than tens of minutes or hours.

I built retrieval pipelines across hundreds of technical PDFs using BM25, PGVector, query expansion, metadata filtering, RRF, and cross-encoder reranking, achieving sub-1.5-second retrieval and roughly 3-second end-to-end responses. I also developed a Neo4j GraphRAG layer covering nearly 1,000 plant and sensor entities, plus a LangGraph multi-agent system with structured outputs, tool calling, PostgreSQL, PGVector, and Neo4j.

Previously at Osense, I developed a RAG system for retrieval and structured extraction from complex documents using Milvus, LangChain, LlamaIndex, and OpenAI. I also trained an NLP classification model to improve document labeling consistency.

My projects include a water-network decision-support system using PyTorch Geometric GNNs and LLMs, and AI assistants for public administration and trading workflows. I'm pursuing a Master's Degree in Artificial Intelligence and Robotics at Sapienza University of Rome.