CV

Education

Stanford University, CA
09/2023 - 06/2025
Master’s in Electrical Engineering
Course: Parallel Computing, Systems for Machine Learning

Cornell University, NY
01/2022 - 05/2022
Exchange | Major in Electrical and Computer Engineering
GPA: 4.24/4.3

Tsinghua University, China
08/2019 - 07/2023
Bachelor’s in Electronic Info Science and Tech
Minor in Statistics
GPA: 3.95/4.0 | Rank: TOP 5

Experience

Apple, Seattle
06/2024 - 09/2024
Machine Learning Engineer Intern

• Developed UI-JEPA, combining a JEPA video encoder with an LLM decoder to generate user intent from videos, outperforming SOTA MLLMs including Claude-3.5-sonnet and GPT-4 Turbo, while using just 5% of their model size, achieving 50.5x lower computational cost, and offering a 6.6x reduction in latency. https://arxiv.org/abs/2409.04081.
• Created the first personalized planning benchmark, MultiCap, evaluating LLM ability to reason over personal context.
• Developed the CAMPHOR collaborative agent framework to enhance LLM responses by actively seeking relevant information and leveraging function call embeddings for retrieval, leading to a 19.5% improvement in response quality, a 30.5% increase in tool accuracy, and a 62.5% increase in plan accuracy compared to baseline models on MultiCap.

Microsoft Research Asia, Beijing
03/2023 - 07/2023
Machine Learning Engineer Intern

• Designed a data compression algorithm using LLMs and LoRA tuning.
• Implemented LoRA tuning and model parallelism for data compression algorithm.
• Explored speculative decoding techniques in causal language models.

SwiftSage
10/2022 - 03/2023
Machine Learning Engineer Intern at 4Paradigm & Research Assistant, Supervised by Prof. Xiang Ren

• Developed SwiftSage, enhancing interactive task completion performance.
• Improved code of ScienceWorld.
• Experimented with SwiftSage, leading to performance boosts.

P-Tuning v2
07/2021 - 01/2022
Research Assistant, Supervised by Prof. Jie Tang, Focus on Natural Language Processing

• Implemented P-Tuning v2, closing the gap for fine-tuning with only 1% tuned parameters.
• Conducted extensive experiments on commonly used pre-trained models.
• Co-authored the paper P-Tuning v2: Prompt Tuning Can Be Comparable to Fine-tuning Universally Across Scales and Tasks.

Open Source Promotion Plan 2023 Project —— DolphinScheduler
08/2023 - 09/2023
Contributor

• Implemented Mixin classes and modularized functionality units.
• Developed code to deploy tasks with additional resource limits like Kubernetes.
• Crafted unit tests, wrote example scripts, and expanded project documentation.

Skills

• Programming Languages: C/C++, Java, Python, R, Verilog, SQL, MATLAB, MIPS Assembly, CUDA C/C++
• Tools: VS Code, Git, PyTorch, Linux, Shell, Docker, Scikit-learn, Pandas, Matplotlib, Azure Cloud