GenAI

multi-modal language models, fine tuning, speculative decoding, reinforcement learning

MRS(modern recommendation systems) - Feed Core ML, ML foundations team

Founding member of Minimal Viable AI. Tech lead for upgrading Feed ranking models to Minimal Viable AI stack. Rewrote trainer, feature proprocessing library, post-training optimization / publishing pipeline, design and implemented debugging / model understanding tools.

2021 - : News Feed ML & Recommendation (MRS)

Founding member of Minimal Viable AI, rebuilding recommendation platform for Feed, Instagram, Reels etc from scratch.

2020 - 2021: AI Platform Infra (Deep Personalization Team, aka DPER)

Next generation recommendation platform for Facebook Ads, Feed, Instagram etc

Supervised by Professor Cynthia Rudin:

1. Single-image super-resolution, published paper at CVPR workshop 2018

2. Interpretable CNN models through concept disentanglement. Paper published on Nature Machine Intelligence

Developed an OCR software for birth records processing

Contribute to PyCLAN library that parses CLAN files