About me

Welcome to my site. My name is Grigoris and I am an Assistant Professor in University of Wisconsin-Madison.

My research focuses on reliable machine learning, focusing on the development of trustworthy models. Concretely,

  • Parsimonious learning: How do deep neural networks learn so effectively using natural signal priors? Natural signals like text and images aren't random; they have a hidden structure, often in the form of sparsity or low-rank constraints. Understanding precisely how networks leverage this inherent structure is a key question. My goal is to design the next generation of networks that have more predictable expressivity (what functions they can represent), trainability (how optimization finds a good solution), and generalization (why they perform well on unseen data).
  • Trustworthy models: How can we build generative models that we can truly trust? Despite their impressive capabilities, existing models can be surprisingly fragile to adversarial attacks, to reliable extrapolations, and can generate nonsensical outputs (hallucinations). My goal is to design the next generation of trustworthy models with predictable and reliable behavior.

News

Funding Acknowledgement

I would like to acknowledge the funding of the following organizations who have generously supported various events or projects in the past. I am very thankful for their support:
  • 2025: Zulip: Sponsored hosting from Zulip, which is an open-source team collaboration tool.
  • 2024: Google and OpenAI: grants on trustworthy Large Language Models (LLMs).
  • 2024: ELISE Fellows Mobility Program: travel grant for short-term visit of an ELLIS lab.