AI and fullstack engineer, based in Milan.
I build LLM systems that can be measured. Most teams shipping AI features change a prompt or swap a model and have no way to tell whether it improved or degraded, so I tend to build the evaluation layer alongside the pipeline rather than after it.
Previously founding engineer at MidFlow, an early-stage workflow-automation startup, where I built the document-understanding pipeline (multimodal extraction with schema-constrained decoding), the LLM evaluation and regression harness, and the cross-platform execution engine.
Now: building an agentic triage pipeline on LangGraph at FOREO.
Python TypeScript · LangGraph LangChain MCP Claude API PyTorch ·
FastAPI Node.js PostgreSQL / pgvector Docker



