Current · 2024 —
Oracle AI CEC
Engineering intelligent systems at the Customer Excellence Center — taking frontier AI from prototype to production at enterprise scale, where reliability is the feature.
AI Engineer · Founder · Est. Singapore
Teaching machines to move through the physical world.
I'm Tim Zhang. I build humanoid robots and the intelligent systems that drive them — turning research into hardware that actually walks, grips, and reasons under real-world constraints.
Role
AI Engineer & Founder
Building at
Oracle AI CEC · Bitbridge Labs
Focus
Embodied AI · Humanoid control
Status
Open to collaborations
01
The premise
Intelligence that can't touch the world is only half-built. For most of the last decade, AI has lived behind glass — predicting tokens, classifying pixels, recommending the next thing to click. The harder, more interesting frontier is embodiment: systems that perceive, decide, and act inside the messy physics of the real world.
That's the work I've chosen. I design the models and the machinery together — perception stacks, control policies, and the humanoid hardware they run on — because the most capable systems are the ones where software and body are shaped by the same hand.
02
Where I build
Founder · Robotics
Bitbridge Labs
An independent lab building humanoid robots from the ground up — actuators, perception, and learned control — with a bias toward hardware that ships rather than demos that don't leave the bench.
03
Humanoid systems
A humanoid is the hardest computer you can build. Every joint is a control problem, every camera frame a perception problem, and the whole thing has to balance on two feet while reasoning about what to do next — in real time, with no second takes.
Software gives a robot its intent. Hardware gives it consequences.
I work the full stack of that problem. From the learned policy that decides where a foot should land, down to the actuator firmware that makes it happen — and back up to the perception models that tell the robot what it's even looking at. The interesting bugs live in the seams between those layers, which is exactly why I refuse to specialise in only one.
04
Selected work
01
Bipedal locomotion policy
02
Dexterous grasp controller
03
On-device perception stack
04
Enterprise LLM pipeline
Swap these for your real projects — each row links wherever you point it.
05
How I think
Build the whole machine. The best robotics doesn't come from a model team throwing weights over the wall to a hardware team. It comes from designing the body and the intelligence as one system — which is the principle Bitbridge is built on.
Ship past the demo. A clip of a robot doing one impressive thing once is a magic trick. Real progress is the boring, repeatable, reliable version — and that's the bar I hold work to.
Stay close to the metal. I'd rather understand a system three layers deeper than I strictly need to than treat any part of the stack as a black box. Embodied intelligence punishes hand-waving.
06