You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I’ve been involved in open source through Google Summer of Code and the Open Source Program Office at UW–Madison.
I built TicketRadar, a ticket discovery platform that received more than 500 waitlist sign-ups and placed top five in the UW–Madison Transcend pitching competition.
I’ve published work in mechanistic interpretability and language-model behavior.
I’m interested in AI safety, reasoning, multi-agent systems, and agent monitoring, with broader interests in evaluations, reinforcement learning, and robotics.
File E-01 Lexon AI SOURCE-GROUNDED AI // MADHACKS 2025 FINALIST
A source-grounded research interface that links generated claims to supporting evidence and makes the evidence structure explorable through an interactive 3D graph.
File E-02 Google Summer of Code 2026 OPEN SOURCE // 7 MERGED PRs
Built an end-to-end benchmarking and ensemble-calibration workflow for PEcAn’s North American carbon reanalysis, culminating in seven merged pull requests across conversion, benchmarking, diagnostics, and regional analysis.
File I-01 A Training-Time Sign Flip During the Formation of an IOI Circuit ICML 2026 // MECHANISTIC INTERPRETABILITY
Traces a training-time reversal in how the repeated-name position affects an IOI model, separating early detection from the later emergence of a suppressor mechanism.
A Training-Time Sign Flip in IOI Circuit Formation : Code, data and analyses for paper accepted at the ICML 2026 Mechanistic Interpretability Workshop.
Lexon AI (MadHacks 2025 finalist) is a source-grounded research interface that generates transparent answers, connects every claim to verifiable evidence, and visualizes the relationships between q...
This repository accelerates anisotropic SOAP descriptors using batched PyTorch operations and Rust. It was developed during Spring 2026 OSPO internship with the Cersonsky Lab at UW–Madison.
Validation, fixes, and optimization improvements for ESSOS, a differentiable stellarator simulation code, done during the UW-Madison OSPO Summer 2026 cohort.
Code, results, and Pythia-160M training checkpoints for "Copying Before Suppression: What Drives a Below-Chance Dip During Language Model Training?" (Findings of EMNLP 2026).