Redmond, Washington, United States
* ML@ Pins, specializing in Recsys * Open source contributor to the “transformers” library * 1st Place Winner of the NYSE x Cohere AI Enterprise AI Hackathon (2024 Feb) * 1st Place Winner of MongoDB x HuggingFace x Nomic AI Gen AI Hackathon (2024 April)
Ranking & Foundational AI (RAI) - Upstream & Foundational Sequence Model Pretraining & Scaling
(See paper attached for GenRec work) Ads Ranking Conversion Modeling - Heavyweight Ranker Modeling & Model Efficiency - Generative Recommender and Foundational Modeling - LLM/LMM in Recys - Model Scaling/Scaling Law - Transformers/Sequence Modeling Optimizations - Offsite Signals Feature Engineering - Deep Learning Based Feature Ranking - Feature Mixers/Feature Interaction - Calibration
Series C startup - Leading NLP & deep learning @ CTO office - LLM supervised fine-tuning, iterative tuning with PPO/KTO/human feedback alignment, retrieval augmented generation - Named Entity Recognition for PII redaction - ML Ops & ML Infra for scalable fine-tuning, batched processing, and real-time inference
Series D startup: Played with LLMs, multimodal models and RAG before they are cool. Founding Engineer for Hive’s AutoML system: training large language models (LLM), multi-task learning text classification, image classification and multimodal models. Helped develop the product and infrastructure from 0 to 1. Performed model and infra level optimizations such as FlashAttention, Mixed precision training, quantization and etc. Hive AI is a series D AI startup founded by 2 Stanford alumnus with over $2 Billion in 2021 valuation, providing a wide range of AI/ML solutions to big customers such as Midjourney, Reddit, Quora, Patreon, BeReal and Onlyfans.
Meeting time recommendation and optimization Most decorated Meta intern by # of achievements badges