What do cross-country rankings and computational biology have in common? A competitive distance runner himself, Assistant Professor Bijan Mazaheri built LACCTiC to compare runners who never race each other — a problem that turned out to require mathematics, engineering, and a hard look at the humans who would read the results. In his postdoc at the Broad Institute, the same problem surfaced again in transcriptomic batch correction. Both are counterfactual questions, answered by a mode of reasoning that today's AI handles poorly. Along the way, Bijan will share his thoughts on engineering within the liberal arts, and why he thinks it may matter more in the age of AI.
This event is free and registration is not required, light pastries & coffee provided while supplies last. No livestream available.
Bijan Mazaheri
Professor Bijan Mazaheri focuses on issues that arise when synthesizing information from multiple datasets, modalities, and batches into machine learning and AI models. He uses tools from theoretical computer science and statistics—including sample complexity, mixture models, and causal inference—to answer questions such as: How can we integrate causal knowledge into statistical models? How can we effectively navigate tradeoffs between data diversity and causality? How can we detect AI-generated or adversarially manipulated data? And, how can we quantify human performance in highly variable environments? Bijan's work is inspired by applications in information security, biology, and human health. He is also an internationally-competitive distance runner and enjoys applying his research to sports analytics.