Stewart Jamieson

I’m an AI researcher and technical leader working on risk-aware, uncertainty-aware machine learning — AI systems that understand the limits of their own knowledge and act safely because of it. As Head of Technology at Themis AI, an MIT CSAIL spinoff, I lead the research and product teams building Capsa: a platform that adds efficient uncertainty quantification to any AI model so that developers can detect and eliminate hallucinations, bias, and failure modes before they reach production.

This work is the direct continuation of a decade-long research thread. My undergraduate thesis at the University of Toronto applied uncertainty quantification to make autonomous-driving vision safer. My doctoral research in the MIT–WHOI Joint Program — co-supervised by Prof. Jonathan How (MIT) and Dr. Yogesh Girdhar (WHOI) — developed risk-aware online learning and Bayesian methods that let autonomous robots reason about the value of information: deciding what is worth learning, communicating, and acting on when operating in unfamiliar, bandwidth-limited environments like the deep sea.

Along the way I’ve been fortunate to have that work recognized — a Best Paper Award at ICRA 2020, and a 2024 article in the journal Artificial Intelligence on finding optimal exploration–exploitation trade-offs through Bayesian risk minimization. Today I’m focused on turning those ideas into reliable, trustworthy AI that organizations can actually depend on.