Exploring how accelerators will be attached to the datacenter network, used for large-scale applications, and programmed with a high-level abstraction, given the increasing compute heterogeneity and system complexity in modern datacenters.

Working at the intersection of biology and architecture, including biologically-inspired computer systems like DNA storage, hybrid electronic-molecular systems, and next-generation bioinformatics applications.

Interdisciplinary machine learning research exploring problems spanning multiple layers of the system stack including deep learning frameworks, specialized hardware for training and inference, new intermediate representations, differentiable programming, and various applications.

Custom databases, hardware, and storage for video applications, like video analytics or VR/AR.

An approximate compiler for C and C++ program based on Clang.

Cross-stack techniques and frameworks to safely trade-off accuracy in computation, storage and communication for better performance and energy efficiency.

In-storage processing architectural exploration for data intensive machine learning, computer vision, and data mining applications

Using Abstract Data Types to expose application-level semantics in datastores

Helping distributed systems programmers reason about and safely trade off consistency for performance

A runtime for providing distributed shared memory (aka PGAS) for large-scale irregular applications, leveraging latency tolerance to overcome the limitations of commodity networks.

Enabling in-network computation in the datacenter network

Enabling large-scale content-based searching by co-designing hardware with visual search algorithms.

Enabling sub-mW face detection and authentication with accelerators near image sensors.

Integrating relational query processing into high-performance parallel languages (e.g., Grappa and other PGAS)

Programming language techniques for automatically designing custom reconfigurable accelerators.

Efficiently supporting multi-quality image storage

Off-the-shelf hardware acceleration for neural networks in the context of approximate computing.

Stochastic circuits and computing applications for low power accelerator and sensor applications.