Hi, I'm Pooja.

About

I am a machine learning engineer, researcher, and writer. My interests are at the intersection of language, speech, and vision. These days, I mostly think about how to make LLMs better at multilingual tasks and more efficient.

I currently work in PyTorch on AI Content Understanding at Meta, where I'm a tech lead for the Machine Translation team. We recently announced multimodal speech translation (voice dubbing) at Meta's Connect conference. You can check out Mark's announcement of our work here.

I started my career as an engineer at Facebook. I was previously a Senior Machine Learning Engineer on the Natural Language Understanding (NLU) team at Facebook Reality Labs. Among other things, I built AutoNLU, a tool for data programming and active learning for NLP. I then became curious about startups and spent some time as a Senior Software Engineer at Impira (acquired by Figma) working on deep learning for document understanding before coming back to Facebook Meta.

I am an M.S. CS student at Stanford in the Honors Coooperative Program (2021-2025), concentrating in Artificial Intelligence. I graduated from the University of Washington in 2017, with a B.S. in Computer Engineering. One of the highlights of my time at UW was working on Respeak, which developed my intersecting interests in NLP, Systems, and HCI. My research received the Best Undergraduate Honors Thesis Award. I also tinkered as a machine learning and software engineering intern at Instagram, Google, and Microsoft.

The best way to reach me is via email.

Personal

I strongly believe in building technology for human good. For health, ethical, and environmental reasons, I am vegan. I also like to read and occasionally write about things I'm learning. My name, पूजा, means prayer in Hindi.