MIT Deep Learning
For the past five years, I have been a lead lecturer and organizer for MIT 6.S191: Introduction to Deep Learning, MIT’s official introductory course on deep learning foundations and applications.
Together with Ava Soleimany, I organize the course from scratch; including developing the curriculum, teaching the lectures, designing software labs, publishing the content online, and handling sponsorship from industrial partners.
We cover a wide range of deep learning methods with applications to machine translation, image recognition, game playing, image generation and more. The course incorporates labs in TensorFlow and peer brainstorming along with lectures. We conclude with project proposals and feedback from the staff and a panel of our industry sponsors.
2021 MIT enrollment of 650 students; enrollment of 300+ students per year in each of 2018, 2019, 2020. Over 30,000 registered students globally and over 5 million lecture views.The entire course is open-sourced and available on http://introtodeeplearning.com.
Research Advising and Mentorship
I am fortunate to advise many amazing undergraduate and masters research students. I was awarded the 2020 MIT Outstanding Mentor Award for Undergraduate Researchers to recognize an individual research mentor for exceptional guidance and teaching in a research setting.
- Jacob Phillips: Unsupervised latent debiasing of time-series models
- Elaheh Ahmadi: Identifying and mitigating the bias in predicting drug development outcomes
- Matt Beveridge: Video-consistent depth estimation for data-driven simulation, joint with I. Gilitschenski
- Alex Knapp: AirGuardian: a parallel autonomy approach to self-flying planes
Undergraduate (B.Sci. - UROP or SuperUROP):
- Selena Zheng: Uncertainty-aware learning of deep neural networks
- Suraj Srinivasan: The effect of temporal processing for vision-based detection
- Alvin Li: Latent space visualization and optimization, joint with W. Schwarting.
- Shinjini Ghosh: Automating the diagnosis of sepsis from only blood-based computer vision
- Julia Moseyko: Depth estimation and reinforcement learning in data-driven simulation
- Catherine Zheng: Model-based RL with transformers, joint with I. Gilitschenski, R. Hasani.
- Roshni Sahoo: Deep orientation uncertainty learning based on a Bingham loss, joint with I. Gilitschenski, W. Schwarting
- Charlie Vorbach: Causal Navigation by Continuous-time NNs, joint with R. Hasani.
- Jacob Phillips: Photorealistic data-driven simulation for autonomous vehicles
- Jordan Docter: Transformers for scalable robot learning, joint with R. Hasani.
- Natasha Maniar: UncOpt: uncertainty-guided optimization for robust learning
- William Chen: Multi-agent autonomous drone policy learning, joint with R. Hasani.
- Diana Voronin: Building an annotation engine for generation of blood-based kymography
- Daniela Velez: Detecting minor fabric defects with machine learning
- Baptiste Bouvier: Attention-based learning with event-based cameras and CT-RNNs
- Tom Dudzik: Data-driven simulation of perception for autonomous vehicles