About

One organisation, two cities — for one reason

The Shenzhen–Hong Kong AI Application Innovation Centre is operated by Jia Heng Technology Development Co., Ltd., registered in Hong Kong and genuinely operating on both sides of the border. We build controllers for embodied robots: the hardware is made in Shenzhen, and the people and the methods come from both cities. We sit on both sides because splitting the work in half kills it — what follows is what each side actually does, and what we want it to become.

Two cities

Shenzhen makes the machine. Hong Kong brings the people and the methods.

"Shenzhen–Hong Kong" is not in the name for decoration. These two things play to two different cities' strengths, and neither substitutes for the other.

Shenzhen

Shenzhen · where things get built

Hardware, engineering, supply chain, validation on real machines

The controller product line. Currently the X-3321 — brain and cerebellum in one box for an embodied robot: twelve automotive-grade cameras, four EtherCAT channels, CAN-FD / RS485 / DIO down to the servos.
Board revisions in weeks. Enclosure, thermals, power, component selection and small-batch production are all local — a hardware revision takes weeks, not quarters.
Bench, compute and real hardware. Whoever comes has a desk, a machine in front of them, and can start the same day.

Why it has to be Shenzhen: the time from schematic to holding the board decides how many iterations fit in a year. That speed exists here and few other places.

Hong Kong

Hong Kong · where people and methods come in

Talent, research, rules and the international channel

Finding people. The motion-control layer needs real depth in real-time systems, fieldbuses and control algorithms. Such people exist in both cities' universities and industry — the Centre approaches them as a Hong Kong entity.
Connecting research. A method from a university only proves out on real hardware. The Centre supplies the machine and the engineering so that step doesn't wait on grant cycles or equipment budgets.
Compliance and reach. A Hong Kong entity, with data handling and contracts under Hong Kong rules — an identity that reads clearly outside the mainland.

Why it has to be Hong Kong: for hiring, for research partnerships and for cross-border agreements, a Hong Kong identity opens doors a Shenzhen one does not.

Neither half works alone. Shenzhen alone builds the machine but cannot reach the people who close the motion-control gap, nor the methods coming out of universities. Hong Kong alone has the people and the methods but no machine you can put your hands on today. Standing on both sides at once is the entire reason the Centre exists.

What we're aiming at

One step at a time, each one verifiable by someone else

Items marked "in progress" have started. Items marked "later" are intentions, not achievements — and we do not write them as though they had already happened.

2026 Q3–Q4

Close the motion-control gapIn progress

The controller hardware is complete; what's missing is the software that makes it move: an EtherCAT master, a real-time control cycle, multi-axis synchronisation, force-position control, planning and safety logic. Two FDE places are open in this first round to build it with us.
2027

Make one real machine finish one real taskLater

Running is not the same as usable. The next step is one specific machine and one specific job — the hardware and software carrying a task someone actually needs from start to finish, with power, latency and failure rates all measured.
Beyond

Turn the combination into something others can useLater

If the first two steps hold, the controller-plus-motion-control combination has a chance of becoming a repeatable product rather than a one-off project. Scale is a conversation for then, not now.

Out of scope

Clear boundaries make for easier conversations

We don't build the machine

Wheeled or legged bases, joint modules, robotic arms — none of these. We build the controller, and the software that makes it move.

We don't run our own research programme

We set no topics, define no research agenda and never touch academic conclusions. The questions come from the people we work with; our job is to have the machine and the engineering hours ready.

No degrees, no admission quota

Enrolment, supervision, academic assessment and graduation standards all stay with the university. A student taking part changes none of them.

We don't take what isn't ours

The method, code and conclusions you produce are yours. We claim no exclusivity over general methodology, and negotiate separately only where the work combines directly with an existing product to form something patentable.

Where we are

Written as it is, including what isn't there yet

Anything a single visit would reveal is better said up front.

Done
The Hong Kong entity really operates. Jia Heng Technology Development Co., Ltd. has been running for over a year with an office, staff and a completed audit — not a shell.
Done
The controller hardware. The X-3321 is in production and can be delivered; the datasheet is available on request.
In progress
The motion-control software layer. Being built, and the reason FDE places are open. This is the Centre's main piece of work right now.
Not yet
A robot to mount it on. We have no base or arm ready to hand over. To validate on a real machine, either you have one or we find one together.
Not yet
A domestic-NPU version. The current controller runs on Jetson; a domestic-NPU version is being ported and is not yet fit to lend out.

Contact

If this is something you'd like to work on

Two FDE places are open in this first round, closing 30 September — and you are welcome to write even if you are not applying. We reply within two weeks.

See what's open lxy@shengangai.com