Bio
Mateo Espinosa Zarlenga is a second-year PhD candidate at the University of Cambridge supervised by Professor Mateja Jamnik and Dr Zohreh Shams. His research, supported by a Gates Cambridge Scholarship, focuses on making AI systems explainable and transparent. Specifically, his work explores the design of interpretable deep learning models that can explain their predictions using human-understandable concepts. Before starting his PhD at Cambridge, Mateo was an early Machine Learning Engineer at Cerebras Systems, an AI hardware startup in Silicon Valley. He holds a BA and MEng in Computer Science from Cornell University and an MPhil in Advanced Computer Science from the University of Cambridge.
Research interests
As deep neural networks become a ubiquitous component in AI systems involved in critical decisions, Mateo's research aims to make these models interpretable and transparent. His work studies how one can construct deep neural architectures that can explain their predictions using human-understandable concepts. Specifically, Mateo is interested in exploring the use and design of such models for real-world setups where concept and label annotations are limited, uncertain, and noisy. At the Turing, Mateo will investigate how encouraging models to construct concept-based explanations can (1) help a human user calibrate their trust in the underlying model, and (2) be exploited at test-time to improve the model's performance when a domain expert is introduced into the loop.