Research

Compute-in-Memory (CIM) Chips

AI, we have a "memory" problem. New computing architectures and chip designs that make more intelligent use of new memory technologies are key to unlocking extreme energy efficiency. We push the frontier of compute-in-memory AI chips through co-designs and prototyping.

Neuro-Symbolic Computing

Neuro-symbolic computing combines human-like symbolic reasoning with hierarchical, fast neural learning, promising more robust, efficient and scalable cognitive AI systems. We work on novel designs and hardware implementations for neuro-symbolic cognitive computing systems. We published one of the earliest hardware demonstrations for vector-symbolic architecture (VSA), also known as hyperdimensional (HD) computing, back in 2016–2018.

CMOS+X Nanokernels and Nanosystems

3D integration of new nanotechnologies and silicon CMOS enables energy- and area-efficient implementations of new compute kernels and system-on-chips. These building blocks can be further packaged into heterogeneous systems for heterogeneous applications.

HPC-at-Edge: Heterogeneous Chiplets

Co-design of heterogeneous silicon and BEOL-augmented silicon chiplets with 3D integration leads towards high-performance computing at the edge.