Peter Belcak

NVIDIA Research

Profile Photo
I'm an AI researcher at NVIDIA. I was previously at Meta, ETH Zurich, and the University of Oxford.

Research

I currently focus on the reliability and efficiency of agentic systems. My past work revolved around efficiency of deep learning models, and the interface of the continuous and the discrete in deep learning, and agent-based financial modelling.

I'm active in professional service: please address reviewing/PC/talk requests at my email. I'm open to collaborations on position, opionion, and survey papers.

Selected Publications

Academic and Public Service

Talks and Panels

Walmart (2025), Salesforce (2025), Julius Baer (2025), Arize AI Deep Papers (2025), HIMSS Panel (2025)

Media Engagements

The work of me and my collaborators has been featured in the Wall Street Journal and the State of AI Report.

Grants Committee

NVIDIA Academic Grants Committee 2025-26

Program Committee

Foundations of Agentic Systems Theory @ AAAI 2026, Multi-Agent Security Workshop @ NeurIPS 2023

Reviewer

The list grows too fast to be maintained here.