Selected Talks

Introducing a new framework for embodied intelligence I call Principle-centric AI: leveraging (and learning) inductive priors to guide learning towards better outcomes (safety, efficiency, robustness).

Why principles might help taking ML to the next level for embodied platforms and how we have successfully used them so far. Invited talk at the ICRA 2022 work on “Fresh Perspectives for Autonomous Driving”

A fun interview where I share my approach to building teams and taking state-of-the-art research from conception to production.

Overview of recent ML Research at TRI across the full robotics stack – (Robust) Perception, (Random) Prediction, and (Risk-aware) Planning – at the RSS 2020 Interaction and Decision-Making in Autonomous-Driving workshop (IDA).

Presentation about our ICRA 2020 paper at the ICRA 2020 Long-term Human Motion Prediction workshop.

Talk given at Wayve on going beyond supervised learning for autonomy. Overview of our recent ICCV and CVPR orals on imitation and perception.

Presentation at Stanford of our work on self-supervised monocular depth estimation (monodepth, a.k.a. pseudo-lidar)

A fun interview I gave in San Francisco at the re:work DL summit in 2019 on open problems in world-scale learning for autonomous driving

Invited talk on our self-supervised learning research, QUVA (joint lab between Qualcomm and the University of Amsterdam)

A talk Chetan Kapoor, Mike Garrison, and I gave in Las Vegas at AWS re:Invent 2018 about our large scale ML infra, one of the first to do distributed deep learning with PyTorch on AWS p3 GPU instances.

An introduction to autonomous vehicles I gave at Hacker Dojo.