Experience

January 2025 — Present · Remote

Under the guidance of Hongru Du, PhD, I’m building an LM for human decision modeling pipeline that generates interpretable decision rules from primitives and iteratively improves them using model feedback. I also incorporate prior probabilities and update them with observed outcomes on a mobility dataset, using gradient descent to better fit parameters and calibrate predicted choice probabilities. Under the guidance of Xinyao Yi, PhD, I’m designing an LLM-driven code optimization benchmark across compute kernels and building an automated harness to propose candidate optimizations, validate correctness, and measure performance/throughput deltas across target platforms.

May 2025 — August 2025 · Seattle, WA

At Amazon, I built and shipped an AWS Lambda service that powers a customer-facing wishlist experience by translating natural-language requests into calls across 20+ internal wishlist/customer-data APIs using LangGraph/LangChain tools. I tackled tricky authentication and added persistent memory plus observability (trace-level logging) to understand how the LLM selected and sequenced tools, looping through LangGraph nodes until it had enough context to answer. I also integrated the service with multiple LLM backends via Amazon Bedrock.

August 2024 — December 2025 · Atlanta, Georgia

At the Automated Algorithm Design Lab, I progressed from a semester-long bootcamp in multi-objective genetic programming and genetic programming fundamentals into the Surrogate Modeling team, where I helped integrate learned surrogate predictors into a neural architecture search pipeline for a computer vision dataset to reduce expensive full model evaluations. After onboarding, I led a small group of students implementing and validating this surrogate in the loop workflow end to end, improving experiment throughput and reducing regressions. Over the following semesters, I focused on developing and tuning graph neural network surrogates, implementing them in the pipeline, running ablations, and iterating on model and configuration choices to better rank candidate architectures and identify high performing configurations with fewer costly training runs.

2025 — 2026 · Atlanta, GA

• (CURRENT) Spring 2026: Studied Generating Functions under Jasper Seabold, covering formal power series, combinatorial enumeration, and analytic techniques • Fall 2025: Studied Quasirandom Graphs under Ruben Ascocii, focusing on pseudorandomness, regularity, and graph properties

January 2025 — May 2025 · Atlanta, Georgia

I contributed to a low-latency market data ingest and trading prototype on an AX7103 FPGA by implementing an ITCH parsing datapath in SystemVerilog alongside a C++ stream generator and testbench for end-to-end validation. I helped build hardware message parsing and pattern-matching blocks and integrated a software harness to replay market data, verify correctness, and speed up iteration. I also contributed shared message formats and constants across the hardware and software boundary to improve interface consistency and reduce integration and debugging time, while researching and evaluating open-source network stacks and hardware-accelerated trading pipeline designs to inform architectural decisions targeting sub-microsecond decision paths.

August 2023 — May 2024 · Atlanta, GA

I researched low-cap cryptocurrency markets, writing scripts to analyze liquidity and price dislocations in order books and identify potential stink bid opportunities during thin trading periods.