About

I am a Member of Technical Staff at Microsoft AI working on post-training and reinforcement learning.

Previously, I was Senior Scientist at Bloomberg. At Bloomberg, I have been building an interactive code agent to help analysts retrieve and analyze the various data provided by Bloomberg. This involves implementing various training procedures to improve large language models (LLMs) on capabilities such as tool calling, and training hybrid retrieval models to improve contextual accuracy in code generation.

I obtained my PhD from Johns Hopkins University, where I was advised by Prof. Benjamin Van Durme. I worked on large language model (LLM) projects including using reinforcement learning (RL) to better orchestrate between LLMs and retrieval systems for code generation [paper], faithfulness evaluation [paper], and preference alignment algorithms [paper, paper]. My research also explored sequence modeling [paper], particularly in the context of streaming, and developed evaluation metrics for code and other structures [paper].

I was awarded outstanding paper awards for Learning to Retrieve Iteratively for In-Context Learning at EMNLP 2024 and Iterative Document-level Information Extraction via Imitation Learning at EACL 2023.