Lijun Zhu Profile Picture

As Director of Computer Vision at Walmart Global Tech, I lead a 150+ person organization building the AI behind Sam's Club — autonomous robots and frictionless exit deployed in 600+ clubs, and a whole-store camera system in pilot. Before that, I grew a data science team from 3 to 20+ and built a demand forecasting platform that improved accuracy by 500+ basis points.

At Texas Instruments, I co-designed binary neural network chips and built model optimization tools and a DL model zoo for edge AI deployment. It all started in research at Georgia Tech, where I applied deep learning to sensor array and embedded/edge devices — with research stints at Microsoft Research and Aramco Research along the way. The thread through it all: taking research ideas and turning them into products that work at scale.

Technical Expertise

AI & Machine Learning
Computer Vision, Time Series Forecasting, Deep Learning, Signal Processing
Platform & Architecture
ML Platform Design, Hardware-Software Co-design, Edge AI Deployment
Leadership
Team Building, Talent Development, Cross-functional Collaboration, Strategic Planning
Languages
Python, SQL, Rust, C/C++, MATLAB
ML Frameworks
PyTorch, TensorFlow, TVM, OpenCV
Infrastructure
AWS, GCP, Kubernetes, Docker

Education

2019
Ph.D. in Electrical Engineering
Thesis: Seismic Processing via Machine Learning for Event Detection and Phase Picking.
Advisor: Prof. James H. McClellan. Center for Energy and Geo-Processing (CeGP).
2013
B.Sc. in Electrical Engineering
Signal processing and machine learning. Capstone: statistical signal processing, detection and estimation in array processing.

Publications

Selected publications (by citation count). See my full list on Google Scholar.

Deep learning for seismic phase detection and picking in the aftershock zone of 2008 Mw7.9 Wenchuan earthquake
Zhu Lijun, Peng Zhigang, McClellan James, Li Chenyu, Yao Dongdong, Li Zefeng, Fang Lihua
arXiv preprint arXiv:1901.06396, 2019
High-resolution seismic event detection using local similarity for Large-N arrays
Li Zefeng, Peng Zhigang, Hollis Dan, Zhu Lijun, McClellan James
Scientific Reports 8(1), pp. 1646, 2018
BCNN: A binary CNN with all matrix ops quantized to 1 bit precision
Redfern Arthur J., Zhu Lijun, Newquist M. Karl
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021

Academic Service

Open Source Contributions

  • PyTorch - Open source machine learning framework
  • NumPy - Fundamental package for scientific computing in Python
  • ObsPy - Python framework for processing seismological data
  • S3I (Seismic Simulation, Survey and Imaging) - MATLAB-based library for seismic simulation and imaging
  • wiggle - Wiggle plot for seismic section data
  • SeisCM - Seismic colormaps for geophysical data visualization
  • ricker - Ricker wavelet generator for seismic synthetic signals