Zixia Hu

About Zixia Hu

Machine learning engineer with 14+ years of experience in large-scale recommendation systems, search ranking, model architectures & infra co-design for efficiency, LLM-based content understanding, and end to end ML system design. Proven track record of delivering high impact production launches, innovating on modeling paradigms, and leading cross functional engineering teams. Strong expertise in deep learning, distributed training, GPU optimization, and statistical learning. Ph.D. in Electrical & Computer Engineering. Author of 1 book, 11 granted patents, and 25+ top tier publications- Machine Learning & Modeling: Large-scale ranking, RecSys, LLM fine tuning, sparse & dense interaction architectures, MUM/BERT, NLP- Systems: Distributed training (FSDP), model/inference optimization (AOTI/lowering), PyTorch, C++- Leadership: Technical strategy, xFN alignment, mentoring, execution across modeling, infra, and product- Mathematical Foundations: Statistical algorithms, numerical optimization, linear algebra.

Experience

1. Staff Machine Learning Engineer

   Meta

   Oct 2024 — Present · San Francisco, CA, US

   Instagram Reels Core Modeling- Model architecture design (feature interaction arch, sequence modeling, dense scaling, sparse scaling, Multi-Task Multi-Label architecture )- Model Infra co-design (Memory/GPU optimization, FSDP - Fully Sharded Data Parallel, AOTI - Ahead of Time Inductor)

Education

• University of Washington

  Phd, Electrical and Computer Engineering

• Delft University of Technology

  Master of Science (MSc), Electrical and Electronics Engineering

• Southeast University

  bachelor, Information Engineering

• Fudan University

  Master

Skills

• Algorithms
• Simulations
• Signal Processing
• Digital Signal Processors
• Matlab
• Wireless
• Ofdm
• System Architecture
• Lte
• 4g
• Cdma
• C++
• Python
• C
• Machine Learning