Sarath Chandar

Timeline

1. 2014

   An autoencoder for bilingual words

   In 2014, he coauthors An Autoencoder Approach to Learning Bilingual Word Representations with Stanislas Lauly, Hugo Larochelle, Mitesh M. Khapra, Balaraman Ravindran, Vikas Raykar and Amrita Saha. The method learns aligned word representations across languages through autoencoders without word-level alignments.

2. 2015–2016

   An IIT Madras master's thesis

   He earns a computer science bachelor's at Anna University and an MS by Research at IIT Madras with Balaraman Ravindran. His 2015 master's thesis, Correlational Neural Networks for Common Representation Learning, wins the 2016 Biswajit Sain Memorial Award for Best MS Thesis in Computer Science.

3. 2018

   Learning with recurrent networks

   In 2018, he coauthors the arXiv preprint Towards Training Recurrent Neural Networks for Lifelong Learning with Shagun Sodhani and Yoshua Bengio. It proposes a curriculum-based benchmark and combines gradient episodic memory with Net2Net to examine sequential learning, forgetting and capacity limits in recurrent networks.

4. 2019

   A PhD thesis on recurrent networks

   Before his PhD, he works for one year as a Research Software Engineer at IBM Research India. He pursues a PhD at Université de Montréal's Department of Computer Science and Operations Research under Yoshua Bengio and Hugo Larochelle; his 2019 thesis is titled On Challenges in Training Recurrent Neural Networks.

5. 2022–2023

   Creating CoLLAs

   He creates the Conference on Lifelong Learning Agents (CoLLAs) in 2022 and serves as program chair in 2022 and 2023.

6. August 2023

   A Canada Research Chair

   In August 2023, he is appointed Canada Research Chair in Lifelong Machine Learning. He leads a research group focused on AI systems that learn and improve throughout their lifetime.

7. 2025

   The NeoBERT encoder

   In 2025, he coauthors NeoBERT: A Next-Generation BERT with Lola Le Breton, Quentin Fournier, Mariam El Mezouar and John X. Morris. The 250-million-parameter encoder combines architectural, data and pretraining advances for bidirectional language models.

8. As of October 2026

   Professor at Polytechnique

   As of October 2026, he is an Associate Professor in Polytechnique Montréal's Department of Computer Engineering and Software Engineering, leads the Chandar Research Lab, and holds the Canada CIFAR AI Chair and the Canada Research Chair in Lifelong Machine Learning. He is a core academic member of Mila and an Adjunct Professor in Université de Montréal's Department of Computer Science and Operations Research. In October 2026, he coauthors RoboJEPA, a world model trained across 12 robotic embodiments to study planning and model scaling.