Pierre Thodoroff

Pierre completed his PhD at ML@CL in 2025, supervised by Neil Lawrence. Before joining ML@CL he was a researcher at the MILA Institute and an intern at IBM Research Israel. His research interests lie at the intersection of theory and practice in Machine Learning, including applications in natural science and healthcare, and theories for data-driven sequential decision making, causality and deep learning. Recent advances in machine learning have been driven by models with significant computational costs. The impact of bounded resources on the algorithm is often ignored and compute allocation decisions are made in an ad-hoc manner by the engineer or researcher.  This can cause issues in deployment requiring real-time inference or cause sub-optimal resource allocation during training. In Pierre’s research, he attempts to incorporate the concept of bounded rationality in machine learning to optimally reason about computations in a data-driven way. He also analyzes the impact of bounded computations on machine learning systems.