Skip to main content

Lijun Zhu

Most AI works in a demo. I build the kind that survives the real world.

Lijun Zhu

I build AI platforms and the organizations that run them. Today I lead a 150+ person engineering and operations organization at Walmart Global Tech building physical AI — models that sense the real world and act in it — with technology live across 600+ Sam's Club stores. Before that, I built the enterprise demand-forecasting platform, and at Texas Instruments I brought neural networks onto edge AI silicon with model optimization tooling.

A brief history

2025-now
Director, Computer Vision & Physical AI
Took a 150+ person organization from a research footing to production delivery, with physical AI live across 600+ Sam's Club stores.
2022-2025
Senior Manager, AI & Data Science
Built the enterprise demand-forecasting platform from two people to 20+ and improved forecast accuracy by 500+ bps.
2019-2022
System Engineer, then Algorithm Team Lead
Brought neural networks onto TI edge AI silicon and led the algorithm team from 2021, after first joining TI as a research intern in 2018.

My work

"Make it work, make it right, make it fast."

— Kent Beck

AI platform · Walmart
Computer Vision Platform
Shared perception infrastructure, co-led with Walmart engineering — data, labeling, models, and edge deployment — so other business units can build on computer vision without starting from zero.
AI platform · Walmart
Forecasting Platform
Time series ML platform serving finance, merchandising, supply chain, and in-store operations — 500+ bps forecast accuracy gain.
Physical AI · Sam's Club
Just Go®
Frictionless exit using computer vision and RFID — deployed in 600+ clubs nationwide.
Physical AI · Sam's Club
Inventory Robots
Autonomous floor-scrubbing robots that scan inventory and detect safety hazards, deployed chain-wide.
Edge AI · Texas Instruments
Model Optimization Toolkit
Pruning and quantization tools (post-training and QAT) for deploying customer models on TI chipsets.
Edge AI · Texas Instruments
DL Model Zoo
Pre-trained deep learning models for common CV, NLP, and ASR tasks optimized for TI hardware.
Edge AI · Texas Instruments
Edge AI Chips
Hardware-software co-design for binary neural network acceleration chips.

How I lead

Structure over heroics. Measurement over demos. A bench of managers who make the calls I would have made. More on how I lead →

Selected writing

All posts →

A bit about me

I teach information systems analysis and design, and business analytics, as an adjunct professor at UT Dallas. Outside work I play clarinet, taught myself piano, and love to BBQ. I make math puzzles for my kids — real math problems disguised as video game scenarios — and run a small server rack at home for self-hosting experiments. This site is one of the things I build for fun.

"An expert is a person who has made all the mistakes that can be made in a very narrow field."

— Niels Bohr