Student, Programmer, Wrestler, Musician

Knowledge Lab: AutoInterp

As a Research Assistant, I optimize Codex-based report generation and analysis for AutoInterp, an autonomous mechanistic interpretability research agentic system.

I also designed an agentic data analysis pipeline for non-technical domains where users can input datasets and query them, producing grounded and reproducible reports.

Research with Dr. Zhewei Sun (TTIC)

My current research focuses on developing NLP techniques to better process and model informal language across languages.

We collected novel datasets for fine-tuning models using BeautifulSoup to more accurately model slang cross-linguistically. We also expanded and modified LLM architectures with PyTorch on distributed GPUs to expand previous research on embedding slang terms.

Prepost AI / Epistemic Hygienist

Engineered a multi-agent system using an AI architecture that integrates NLI nodes and RAG retrieval to mitigate "bullshit" assertions and promote nuanced social media speech.

The system gathers external context from sources like Wikipedia and BBC to perform structured reasoning. I validated performance across diverse agentic designs using models like Gemma 3, Llama 3.1, and GPT-OSS.

Isolated Knowledge Updates in LLMs

Quantified the limits of localized model editing using arithmetic contrapositives on the Qwen2.5-1.5B model.

Implemented dynamic control over the edit-preservation tradeoff using an inference-time "alpha sweep" on LORA task vectors, accomplishing 100% edit efficacy. Successfully isolated targeted factual updates using a 50-item multi-domain locality set.

*Note: The associated repository is currently private.

Nabu - Personal AI Language Tutor

Developed an AI native desktop application for real-time conversations with a personalized AI system to expand vocabulary using spaced repetition.

Crafted the workflows, prompts, and API endpoints, implementing a host of AI tools and APIs such as LangGraph, Whisper, and ElevenLabs.