CV

Education

• M.S. Electrical and Computer Engineering, University of California, Los Angeles (UCLA), Sep 2024 – Dec 2025
• B.S. Computer Engineering, University of California, Los Angeles (UCLA), Sep 2021 – Jun 2024

Research Interests

Agent, Reinforcement Learning, Generative Models (Diffusion, LLM).

Research Experience

• Graduate Researcher, AGI Lab (Apr 2024 – Present)
  • Advisor: Prof. Quanquan Gu, UCLA, Los Angeles, CA
  • Developing Linear Maillard Sampling for linear bandits: implemented a linear ridge-regression parameter estimator and a Boltzmann exploration mechanism.
  • Conducting theoretical regret analysis using self-normalized concentration arguments and elliptical potential lemma; drafting theorem proof notes and preparing a manuscript.
  • Reproduced diffusion-based energy-guided sampling baselines for offline RL with large-scale experiments on computing clusters, and evaluated generation quality with FID on ImageNet.
  • Lead development of the EurekaClaw project (GitHub): designed and implemented the frontend interface, coordinated open-source community contributions, and maintained codebase for research and public use.
• Undergraduate Researcher (Sep 2022 – Dec 2023)
  • Advisor: Prof. Lin Yang, UCLA, Los Angeles, CA
  • Model-Based RL with Generative Model Policy Learning: Studied perturbed model-based planning for discounted MDPs; re-derived core lemmas and compiled theorem-level proof notes.
  • Zero-Sum Markov Games with Boltzmann Exploration: Investigated extending POLITEX mixture policy iteration to simultaneous-move two-player zero-sum Markov games in an online learning setting.
  • Local Sampling Access in Discounted MDPs: Formalized a practical sampling method that permits queries at previously visited states with resets.

Research Projects

• Web World Models (WWM) – Princeton AI Lab, Prof. Mengdi Wang (Dec 2025)
  • Prototyped a Web World Model framework that generates unlimited context via LLMs in text/code environments without data storage.
  • Built multiple WWM-style environments spanning real-world and fictional settings, demonstrating controllable and open-ended interaction for users and agents.
• TENG++: Time-Evolving Natural Gradient for Boundary-Enhanced PINNs – Advisor: Prof. Di Luo (UCLA)
  • Extended the TENG framework with Dirichlet boundary constraints in PINNs; implemented a Jacobian-based natural-gradient least-squares stepper and integrated it with Euler/Heun time-stepping.
  • Validated on heat-equation benchmarks with analytical solutions. (GitHub)
• Hallucination Detection and Evaluation of LLMs – Advisor: Prof. Yuan Tian (UCLA)
  • Built a KnowHalu-style hallucination evaluation pipeline: Replaced KnowHalu’s judge with HHEM to accelerate evaluation while maintaining QA detection performance (76.9% accuracy); explored a non-fabrication-checking variant reaching 82.2% accuracy.
  • Improved summarization hallucination detection via segment-level verification.
• Predicting Text from Intracranial Neural Signals – Advisor: Prof. Jonathan Kao (UCLA)
  • Implemented RoPE positional embeddings in a CNN-Transformer decoder for intracortical speech-to-text; implemented the CR-CTC loss, reducing CER from 0.234 to 0.1636. (GitHub)
• Deep RL Algorithms in Cryptocurrency Trading – Advisor: Prof. Jonathan Kao (UCLA)
  • Built an ETH trading environment with trading features and actions through hourly Bitfinex ETH dataset.
  • Re-implemented off-policy actor-critic agents (DDPG, TD3, SAC); found SAC achieved the best returns while TD3 trained more stably. (GitHub)

Selected Coursework

Stochastic Processes; Information Theory; High-Dimensional Statistics; Statistical Machine Learning; Theory of Reinforcement and Online Learning; Convex Optimization; Large-Scale Optimization Theory; Applied Deep Learning.

Teaching Experience

• Teaching Assistant:
  • ECE 132B: Data Communications and Telecommunication Networks (Prof. Izhak Rubin, Jan 2025 – Apr 2025)
  • ECE 241: Stochastic Process (Prof. Lin Yang, Jan 2025 – Apr 2025)
• Learning Assistant:
  • ENGR 96A: Machine Learning With Python (Prof. Jacob Schmidt, Sep 2023 – Jun 2024)
• Reader:
  • CS 51A: Logic Design of Digital Systems; ECE 131A: Probability Theory; CS M146: Machine Learning

Awards & Honors

• Awards: Full Tuition Grant (Fall 2024, Winter 2025); AWS Computing Fund for BigANN Competition (NeurIPS 2023)
• Honors: Dean’s List (Spring 2023); Eta Kappa Nu (HKN) Honor Society (Jun 2022); Upsilon Pi Epsilon (UPE) Honor Society (Jun 2022)
• Social: RSS 2025; NeurIPS 2025

Skills

• Languages: Chinese (native), English (proficient), Spanish (principiante)
• Tools: C/C++, Python, MATLAB, LaTeX, GCP, Docker, Conda, Git, Bash Scripting, Tmux
• ML: JAX, PyTorch, Weights & Biases, Huggingface/Transformers, NumPy, Matplotlib, Scikit-learn, Pandas, Gymnasium, VeRL