Michelle Z. He
Hi, I'm Michelle Z. He, a passionate and dedicated computer science graduate currently pursuing my Master of Science at the University of California, San Diego.
About
Hi, I’m Michelle Z. He, a passionate and dedicated computer science graduate currently pursuing my Master of Science at the University of California, San Diego. I specialize in video coding algorithms and software engineering with hands-on experience in advanced video compression standards, deep learning model tuning, and full-stack web development. I’m driven by solving complex algorithmic challenges and creating efficient, scalable solutions.
With experience working at Nokia and Shanghai Fullhan Microelectronics, alongside research contributions at Shanghai Jiao Tong University, I continuously seek to enhance my technical expertise in AI and video codecs. I thrive in collaborative environments, embrace Agile methodologies, and enjoy leveraging modern tools and frameworks to deliver impactful projects that blend innovation with real-world application.
- 3 years' experience
- Joined August 2025
Industry experience
- Software & Internet
- Computers & Electronics
- Professional Services
- Education
More experts in San Diego, United States
Work experience
Designed novel video compression algorithms in C++ for JVET's next-generation video coding standards. Implemented EIP filters using diagonal pixel samples, improving luma bitrate by -0.02% (AI) and -0.01% (RA), with contributions accepted at JVET meetings. Enhanced temporal partitioning prediction algorithms by leveraging same-QP frames in reference lists instead of collocated frames, improving codec efficiency and performance.
Mastered H.265/HEVC and H.266/VVC standards and implemented a C++ algorithm enabling early block partition prediction using spatial features like edge complexity and smoothness. Optimized encoding process by enforcing structured block sizes and early exit strategies, achieving a 68% reduction in encoding times with only a 3.5% BD-rate increase, balancing speed and quality effectively.
Supported the project on Out-of-Domain Human Mesh Reconstruction via Dynamic Bilevel Online Adaptation, including building a Blender-based rendering pipeline for results from streaming videos. Tuned deep learning models using Python and evaluation metrics like MPJPE/PA-MPJPE to improve out-of-domain generalization for 3D human mesh reconstruction tasks, contributing to research published in IEEE TPAMI.
Education
Hire Michelle Z. He today
Post a job and invite Michelle, or contact them directly. No hidden commission.