Peter Izsak

I’m an applied research scientist at Microsoft, working on natural language processing for healthcare. My current work focuses on agentic clinical systems, retrieval-augmented generation, and evaluation of generative and agentic AI.

Previously, I was a research scientist and team lead at Intel Labs. I hold an M.Sc. and B.Sc. from the Technion, Israel Institute of Technology.

My published research examines how to train and compress language models with fewer resources, improve retrieval, and evaluate long-context models. This includes papers at EMNLP, ICLR, COLING, and SIGIR, and NeurIPS workshops. Earlier work covered information retrieval and distributed systems.