Soumith Chintala

AI Researcher, Engineer, Community Builder

I am an Artificial Intelligence researcher, engineer and community builder.

I am currently at Meta, jumping between Engineering, Research and Leadership as I find convenient. I also visit NYU as a part-time researcher.

My career interests have been defined by two sets of work: AI Platforms/Ecosystems and AI Research. I’ve also spent enough time leading large orgs that I have all kinds of (what I think are) crazy ideas about org-design.

AI Platforms / Ecosystems

I love open-source communities and advocate for open research.
I think deeply about the cultural aspects and social dynamics of open-source software.
I advocate for a simplistic style of programming, and I like fast prototyping of ideas, sometimes using unconventional engineering practices.
I think products and communities have to be rapidly iterated and evolved to get to a good state – with feedback loops and incentive structures designed very carefully to get to good dynamics.

In this vein, I’ve helped build the machine learning platform PyTorch and maintained Torch-7, EBLearn and several other open-source projects in the machine learning domain.

  • Torch-7 at it’s peak was used by Google Deepmind, Twitter and Meta and singificant AI research was powered by it.
  • PyTorch currently powers most of the world’s AI research and product – with hundreds of companies, research labs and individuals using and maintaining it. It has significant and tangible real-world impact from self-driving cars to drug discovery to cancer research to NASA’s Mars Rover to several consumer products, and that amount of real-world usage often intimidates me. There’s more than one bug in PyTorch, when it was uncovered, I couldn’t sleep thinking about what the downstream effects could be.

To deeply understand the products I build, I like helping users with their problems simple or complex, and have answered thousands of questions across the PyTorch and Torch forums, investing a significant portion of my professional career in this endeavor. These signals shape our own understanding of products deeply in important ways. I learnt more about building the right ML products from this exercise than from anything else.

I learned interesting things while maintaining Torch, which was written in Lua – a large project written in a niche language. If you ask me about it, I’ll tell you fun stories.

AI Research

My current primary interest is to build a household robot that helps me with all kinds of chores. To help this robot reason well with little data, I want to build a world simulator (so that it can rollout scenarios in it’s head and pick the best ones). To build this world simulator, I’ve been interested in multi-modal models (that combine vision, speech, text, robots), generative models (for vision and speech) and efficient representations for encoding the human-centric world. My secondary interests are in research that intersects between Computer Systems and Machine Learning, partly motivated by my work on Machine Learning Platforms.

I’ve recently made a decent amount of progress on home robotics at NYU, with my collaborator Lerrel Pinto, often using a Hello Robot | Stretch. Here’s some of my recent work that I’m excited about:

I often try to avoid toy or hypothetical problems, even as proxies and ground my research towards applications with obvious benefits. I am not a theoretician for the lack of aptitude or focus, I respect those who make progress in theory.

In the past, I’ve worked on robotics, object and human detection, generative modeling of {images, videos}, AI for video games, ML systems research. Some of my most cited work is on Generative Adversarial networks (GANs), where I co-authored three well-cited papers: LAPGAN (demo), DCGAN (code/demo) and Wasserstein GAN.

You can see a full list of my peer-reviewed or pre-print manuscripts on my Google Scholar page.