Shan Wang

What they work on

1. Improve the thermal robustness, switching speed, endurance and energy efficiency of MRAM, a memory that stores bits in nanoscale magnets, by tackling nanofabrication and design challenges for mobile AI.
2. Combine MRAM-based associative memories, which retrieve stored data by similarity, with embeddings-based neural networks so phones and watches can run personalized AI that keeps private data on the device.
3. Fabricate a 2-terminal spin-orbit torque MRAM using materials with unusual spin behavior, such as MnPd3, to reach unlimited endurance, nanosecond writes and switching without an external magnetic field.
4. Design and manufacture a processing-in-memory chip that performs sparse multiply-and-accumulate operations inside the memory, supporting both running and training neural networks.
5. Build SAS-MRAM, which combines spin-transfer and spin-orbit torque writing, on top of a TN40G CMOS chip using a custom back-end process, with micromagnetic and HSpice simulations to create design kits.
6. Develop continual learning algorithms that let edge AI chips learn new, user-specific data without forgetting earlier knowledge while minimizing energy-costly memory writes.

More about the research

Why it matters

Smartphones, smartwatches, robots, cars and connected health devices increasingly rely on AI. Running that AI locally drains batteries and is limited by power-hungry memory. Nonvolatile magnetic memory that uses no standby power could let these devices learn from personal data on the device itself. This could extend battery life, keep sensitive data private, and cut the energy cost of AI in applications such as autonomous driving and speech assistants.

Known grants

From NIH and NSF research grants. Labs often have other funding we can't see.

Closest labs

The school

Stanford University: 87 labs on PI Pages, $1.6B a year on research, 739 research doctorates a year.

More about Stanford University

Reach out

A start for your email. Fill in the brackets.