I am a computer science undergraduate at UC Berkeley and a Large Language Model Engineering Intern at Douyin AI, ByteDance. My work focuses on agent self-evolution, code intelligence, AI-agent evaluation, and reliable long-horizon systems.

Resume PDF | Google Scholar | GitHub

Research Interests

• Agent self-evolution and automated improvement
• Code intelligence and software-engineering agents
• AI-agent evaluation
• Reinforcement learning and post-training
• Reliable long-horizon agents

Featured Work

• Simple Evolve Agent — Coming soon. Infrastructure for reproducible agent self-evolution research, supporting iterative improvement with sandboxed evaluation, provenance tracking, and reliability guardrails.
• EffiSkill: Agent Skill Based Automated Code Efficiency Optimization — Under review at EACL 2027. First-author work on automated code-efficiency optimization using reusable agent skills mined from slow and optimized program pairs.

Experience

• Large Language Model Engineering Intern, Douyin AI, ByteDance — May 2026–Present: working on agent self-evolution, practical agent systems, and evaluation for long-horizon and code-oriented tasks.
• Researcher, Shanghai Jiao Tong University (Spring 2026): built EffiSkill, scalable inference pipelines, and offline evaluation workflows on EffiBench-X for code-efficiency optimization.
• Research Assistant, UC Berkeley (Fall 2025): developed data-cleaning and analysis pipelines on a 1.2M-row gig-economy dataset and supported collaborator-facing empirical analysis.
• Researcher, Shanghai Jiao Tong University (Summer 2025): implemented SFT + GRPO training pipelines for code LLMs and ran large-scale ablations on 8 x A100 GPUs.
• Teaching Assistant, UC Berkeley (Fall 2024): mentored students on Data 8 for data science and Python programming.