Hello.

My name is Matteo. I research and build ways to make AI systems fail. I lead R&D at Giskard.

My work

— I'm CTO at Giskard AI, a European startup building tools for AI evaluation and safety. Some things I've worked on:

Phare, a multilingual LLM safety benchmark (initially developed in collab with Google DeepMind)
Red Teaming LLM Applications, a course with Andrew Ng at DeepLearning.AI
Adversarial ML CTF at DEFCON 31
Open-source tools for automated AI evaluation (Giskard Scan, 5k+ GitHub stars)

Previously: PhD in applied mathematics at ENS Paris, MBA, software engineer, cofounder of a medical AI startup.

Open hours

— I keep a few slots open each week for conversations. No agenda needed. Book one here.

Links

— The usual suspects: