Hugo Larochelle

Hugo Larochelle is a deep learning researcher who has been scientific director of Mila, Montréal's AI institute, since September 2025, a post once held by his doctoral supervisor Yoshua Bengio. His 2008 paper "Zero-data learning of new tasks" introduced the concept of zero-shot learning, and after a postdoc with Geoffrey Hinton in Toronto he coauthored the NADE and MADE autoregressive models in 2011 and 2015. While a professor at the Université de Sherbrooke from 2011 to 2016, he co-founded the machine learning startup Whetlab, which Twitter bought in 2015. Google recruited him in November 2016 to lead a Montreal arm of Google Brain, and he went on to serve as general chair of NeurIPS 2020 and as a founding editor-in-chief of Transactions on Machine Learning Research from 2022 to 2025.

Timeline

By 2008

Doctoral work with Yoshua Bengio

At the Université de Montréal, he earns a BS in mathematics and computer science and an MSc in computer science, then pursues a PhD with Yoshua Bengio as his supervisor. In 2008, he coauthors "Zero-data learning of new tasks" with Dumitru Erhan and Bengio, which introduces the concept of zero-shot learning, and a denoising autoencoder paper with Pascal Vincent, Bengio and Pierre-Antoine Manzagol.
2011–2015

Autoregressive models NADE and MADE

After his PhD, he does a postdoc with Geoffrey Hinton at the University of Toronto. In 2011, he and Iain Murray publish the Neural Autoregressive Distribution Estimator (NADE) at AISTATS, where he receives a Notable Paper Award in 2011; in 2015, MADE, a masked autoencoder for distribution estimation written with Mathieu Germain, Karol Gregor and Murray, appears at ICML. These models help popularize the neural autoregressive approach now common in generative AI.
2011–2016

Professor at Sherbrooke

From 2011 to 2016, he is a professor in the computer science department of the Université de Sherbrooke's Faculty of Science, working on machine learning. With Jasper Snoek and Ryan P. Adams, he coauthors "Practical Bayesian Optimization of Machine Learning Algorithms" in 2012, an automatic approach to tuning a learning algorithm's hyperparameters.
June 2015

Whetlab joins Twitter

His research at Sherbrooke leads him to co-found Whetlab, a startup whose technology makes machine learning better and faster for companies. Twitter announces the acquisition in June 2015, and he goes on to work in Twitter's Cortex group.
November 2016

Leading Google Brain in Montreal

In November 2016, Google opens a deep learning research group in Montreal as a remote arm of its Google Brain team, and he returns home from Twitter in Boston to lead it. The lab later becomes part of Google DeepMind, where he serves as a principal scientist in Montreal.
2018–2020

CIFAR chair and NeurIPS 2020

CIFAR appoints him a Canada CIFAR AI Chair in 2018, and he is an associate fellow in its Learning in Machines & Brains program. He serves as general chair of the NeurIPS 2020 conference while at Google Research.
2022–2025

Founding editor of TMLR

He co-founds the journal Transactions on Machine Learning Research and serves as one of its founding editors-in-chief from 2022 to 2025, with Kyunghyun Cho and Raia Hadsell. He is the last founding editor-in-chief to complete his term, and Laurent Charlin joins the editors-in-chief in January 2026.
Since September 2025

Scientific director of Mila

In September 2025, he becomes scientific director of Mila, Montréal's AI institute, a role once held by his former supervisor Yoshua Bengio; he chairs its scientific council and leads its research direction. He is also an adjunct professor at the Université de Montréal and McGill University and scientific lead at Adaption Labs. In 2026, he gives a talk at the Université de Sherbrooke's Institut quantique on applying AI to AI research.