Projects & Research

Small Language Models for Edge AI in Space

Ongoing project with UC Berkeley and Satlyt, focusing on building an AI agent workflow with tool-calling capabilities for satellite edge computing.

AgentProbe

Built a red-teaming framework for AI agent systems deploying a four-agent adversarial swarm.

Enterprise Nervous System

Cross-silo incident-resolution swarm that queries logs, GitHub, and Jira to automate root-cause analysis.

CodeBase Agent

Agentic AI code refactoring system using a multi-agent pipeline that analyzes Python code and validates changes.

Smooth Cascade Unlearning through Reversed Self-Distillation

Researched in-context unlearning and parameter-efficient fine-tuning to make LLMs forget private data.

Data Analysis for Table Tennis Matches

Conducted a study on table tennis ball trajectories, utilizing physics-based modeling and player data to analyze bounce uncertainty zones and strike timing, contributing to player classification and strategy development.

Type System for Secure Information Flow

Implemented a type system and checker in OCaml to ensure the noninterference property.