I build LLM systems that read documents, and I measure whether they work.

I am an AI engineer. Since 2023 I have been at Value Partners, where I started as an intern and now lead a team of four on AI projects in the legal and insurance sectors — architecture, estimates, mentoring — while still prototyping myself.

The thread running through almost everything I have built is LLM systems applied to complex documents: structured extraction from legal filings and medical paperwork, hybrid RAG, chatbots, agents. Legal, insurance, finance, credit management, automotive. Some of those systems are in production, some stopped at proof of concept, and one reached production and was then barely used — all three happen, and saying so is more useful than only telling you about the first.

If one thing follows me from project to project, it is measurement. In almost all of them I built the evaluation framework alongside the system: precision, recall, F1, exact and partial match, a calibrated LLM-as-a-judge, annotated ground truth. Not for the metric itself, but because it is the only thing that separates a real improvement from a regression once you start changing prompts. It is also the subject of the talk I gave at DevFest Pisa 2026, “LLM Testing 101”.

I hold a BSc in Computer Science from the University of Pisa — the thesis was on data extraction in NLP projects — and the EU AI Act Foundation certification.

You will find my articles here, in Italian for now.

How to reach me

The reliable way to write to me is LinkedIn — the profile is just below, among the links. I read everything; I reply to the things I have a reply to.