| layout |
|---|
default |
Hi! I’m Jacob, a Masters' student at Stanford University studying Artificial Intelligence. My academic interests are broad but roughly center on understanding ML systems through their applications. Among my many hobbies, I enjoy sports, cooking, and D&D podcasts.
Here are links to my resume, github, and linkedin. You can leave me anonymous feedback here.
This past summer (2024), I was a Visiting Researcher at the Alignment Research Center, where I formalized a surprise account of the Linear Argmax model. You can find the writeup here.
In 2023, I was a Research Fellow at the Existential Risks Alliance, where I investigated Contrast-Consistent Search as a detector of LLM Truthfulness under the mentorship of Tilman Rauker. I've since continued and expanded this research direction at Stanford under the mentorship of Prof. Vasilis Syrgkanis.
In 2022, I interned with Jane Street and played figgie all day used a diverse software suite (python, excel, bloomberg, SQL) to research and develop trading strategies.
In 2021, I interned with Five Rings Capital, researching under the mentorship of a full-time quant and competing in proprietary training games.
In 2020, I worked at the Openproof Foundation, developing a node.js frontend for mathematical logic courseware.
In 2019, I interned with ASR Analytics, developing a graph-based recommender engine to inform IRS tax compliance interventions.
In 2018, I participated in the University of Maryland REU Program, researching a finite graph approach to the Hadwiger-Nelson problem. The resulting paper was published in the journal Geombinatorics.
During Winter and Spring 2021, I researched the Outer AI Alignment problem under the mentorship of Alex Turner, extending Turner's definition of POWER to multi-agent games. I presented at the Winter and Spring SERI conferences and coauthored several blog posts as informal writeups.
During Winter 2019-20, I studied Algebraic Topology (Hatcher) under the mentorship of Joseph Helfer. I also gave a talk presenting the topic to undergraduate students (slides).
As a MBMBAM enjoyer, I have dearly missed the departed Yahoo! Answers QA service. To remedy this, I fine-tuned a language model to simulate the answers submitted to the podcast.
To help my friends in a discord server set personal goals, I coded a bot to let users set and evaluate time-bound goals.
In an effort to teach myself Reinforcement Learning techniques through game-playing, I wrote a general framework for algorithmic multiplayer game-playing. Alice (my alpha-beta pruning algorithm with clever tweaks, named in honor of cryptographic legends Alice and Bob) can tie me at tic-tac-toe, but not at chess.
During high school (Montgomery Blair HS), I was a member and later co-captain of the Blair Math Team. During that time, I wrote up lecture notes on various math topics, with a specific focus on audience participation and making learning fun. Some highlights are listed below:
During Fall 2020, I worked with fellow Magnet alumni Laura Cui and Steven Qu to design and take a survey of students' experiences in the STEM Magnet program. Our results are available here, or see the Magnet Foundation press release.
- A D&D campaign module, designed for storytelling in the style of TAZ Balance
- Meta-chess, a team-based strategy game designed for a group of chess enthusiasts. I planned to organize a game at Mathcamp but never got around to it :(