Math · CS (AI) @ Stanford · Builder
About
AI tools, research,
and a consumer app.
— math + CS (AI) at Stanford.
I've been writing code inside research labs since high school — first at the Institute for Systems Biology (computational biology), then through college at NASA JPL's Mars Science Laboratory, Stanford's GPS Lab, and Bridgewater.
Some things from those internships: a 90% speedup on Curiosity rover daily planning at JPL; GPS signal authentication strengthened to survive 40% packet loss at the GPS Lab; two systematic trading models at Bridgewater (a long-rate bond model and a Chinese equity flow model), plus a natural-language financial-research tool at the hackathon.
Currently building an adaptive AI language tutor — closed-loop conversation engine, sub-second latency, context-aware FSRS scheduling, and phoneme-level pronunciation feedback. Some research on the side (DPO alignment, gravitational lensing detection in CV), and the occasional oil painting.
Work Experience
- 01
Investment Engineer Intern
Bridgewater Associates · Stamford, CT
Built and backtested two systematic trading models: a long-rate forecast against bond futures, driven by growth and inflation signals; and a Chinese equity flow-prediction model. Also built a natural-language financial research tool at the Bridgewater hackathon.
Jun — Aug 2025Details → - 02
SWE & Research Intern
Stanford GPS Lab · Stanford, CA
Rebuilt a GPS authentication system in MATLAB with stronger cryptographic guarantees (HMAC 16 → 28–37 bits, aggregated auth). Implemented error-correction tolerating up to 40% message loss. Reached 100% test coverage and containerized the dev environment.
Jun — Sep 2024Details → - 03
MSL Robotics & SWE Intern
NASA Jet Propulsion Laboratory · Pasadena, CA
Accelerated Curiosity rover planning scripts by 90% via modular refactoring, mocked service calls, and runtime profiling. Built an automated HTML reporting pipeline for rover operations with dynamic text and image generation.
Jun — Aug 2023Details → - 04
Computational Biology & Data Science Intern
Institute of Systems Biology · Seattle, WA
Computational analysis of microbiome data from the American Gut Project — used Python, NumPy, and Pandas to surface links between lifestyle factors and microbial diversity.
Jun — Aug 2021Details →
Projects
Metaphor — Japanese Conversation Tutor
Japanese conversation tutor for self-directed adult learners. Vocab apps drill words but don't make you conversational, raw LLM chat doesn't remember what you know, and human tutors cost $50/hr — Metaphor gives you scenario-driven speaking practice on demand, with a durable learner profile that tracks per-element proficiency across sessions.
Disincentivizing RL Agents From Hiding Reward Hacking
Alignment research showing that RL agents trained on proxy rewards have a structural incentive to hide reward hacking from RLHF evaluators — and demonstrating standard deep RL agents act on that incentive, behaving correctly when observed and reverting to proxy exploitation when not.
Optimal Language-Learning Curriculum Design via MCTS
Formalized language-learning curriculum design as an MDP and solved it with Monte Carlo Tree Search. On a 105-item Japanese (JLPT N5–N4) library evaluated against 1000 simulated heterogeneous learners, MCTS reached conversational fluency 36.2% faster than random baselines (p < 0.001) while maintaining 25.9% higher retention.
Direct Preference Optimization for GPT-2 Downstream Tasks
Implemented GPT-2 from scratch, then fine-tuned with Direct Preference Optimization on sentiment analysis, paraphrase detection, and Shakespearean sonnet generation. DPO consistently improved paraphrase and sonnet quality versus standard maximum-likelihood fine-tuning while sidestepping the instability and overhead of RLHF.
Contact
Say hello
— End of transmission · X91