I turn complex models into practical, real-world systems — scalable generative AI applications and low-latency explainable AI that build user trust and drive business impact.
I genuinely enjoy the hard puzzles: making advanced AI both highly performant and completely transparent. BSE in CS from Princeton (magna cum laude), MEng in CS from Cornell, AWS Certified Developer.
Driving an internal agentic AI assistant that helps 4,000+ employees retrieve enterprise knowledge through natural language. Own a workstream spanning agent orchestration, RAG pipelines, evaluation and guardrails, and low-latency backend services.
LLM-powered commerce search that turns natural-language prompts into automated cart builds, scaling to 1M+ transactions a year. Engineered a low-latency explainable-AI framework giving enterprise customers real-time, profile-based justifications for recommendations. Co-authored APLE (WebSci ’26).
COS 445 Economics and Computation — auction theory, market design, computational social choice.
Automated performance checks for backbone network devices and built visualization tools tying together multiple databases.
Computer-vision system coordinating autonomous robots over 5G and edge computing — multi-object tracking, path prediction and velocity correction on AWS.
Analyzed spatial data from VR simulations with Python and R to inform gaze-contingent display design.
Neural causal models and pre-generated templates produce faithful, personalized recommender explanations in real time — ~45 ms latency, 74% faithfulness, no changes to deployed models.
A detector for diffusion-generated imagery to counter misinformation, with a qualitative study of the attributes that give synthetic images away.
Tracks multiple robots and adjusts their velocities autonomously — a demo of low-latency 5G edge computing with AWS in the automotive space.
A causal framework for detecting comment-level implicit gender bias, showing how detection differs across social platforms.