CV

Graduate Research Assistant

University of Southern California

• Developed EPSVec (ICML 2026), a method for efficient and private synthetic data generation via representation engineering.
• Work on the foundations of trustworthy AI, including the decision-making capabilities of LLM agents and alignment-focused sampling mechanisms.

Research Intern

Sooth Labs

• Hosted by Dr. Russ Salakhutdinov and Dr. Yaser Sheikh.
• Developing a state-of-the-art agent harness for long-horizon event forecasting.

Undergraduate Research Assistant

University of Chicago

• Led the development of Prophet Arena (ICLR 2026), a live prediction benchmark evaluating LLMs’ ability to retrieve and aggregate evidence for probabilistic forecasting.
• Featured by The Atlantic, Yahoo Finance, and The Dispatch.

Software Development Intern

Amazon Web Services (AWS)

• Built an account resource termination pipeline for AWS accounts with fraudulent activity, saving $11M in the first six months of deployment.
• Implemented a proactive security module for account operations with multi-stage integration tests; received a return offer.

Ph.D.

University of Southern California

Computer Science

• Advised by Prof. Sai Praneeth Karimireddy

M.S.

University of Chicago

Statistics

• Advised by Prof. Haifeng Xu
• Advanced Scholars Program

B.S.

University of Chicago

Computer Science (with Honors), Statistics

• Cum Laude

LLM-as-a-Prophet: Understanding Predictive Intelligence with Prophet Arena

EPSVec: Efficient and Private Synthetic Data Generation via Data Vectors

Sampling More, Getting Less: Calibration is the Diversity Bottleneck in LLMs

Psychological Steering via Calibrated Activation Injections