I am an assistant professor in the Department of Computer Science and Electrical and Computer Engineering at Duke University. My research focuses on harnessing the unprecedented computational power and efficiency of hardware acceleration to expedite and advance research in areas related to big data, the natural sciences, healthcare, and artificial intelligence, ultimately improving human flourishing. Specifically, leverage the insights derived from a deep understanding of applications (e.g., data structures, data access patterns, computational patterns, application characteristics) to create innovative hardware and software systems to tackle impactful problems. In addition, I am improving the ecosystem to enable architects to develop and deploy accelerators more easily, and to allow non-hardware users to leverage the efficient accelerated systems with a low barrier to entry.
Thank you Google for the generous gift to support our research! We will be using this award to explore hardware and software methods to accelerate vector databases and retrieval-augmented generation for large language models. Trinity College of A...
As the parameter size of large language models (LLMs) continues to expand, the need for a large memory footprint and high communication bandwidth have become significant bottlenecks for the training and inference of LLMs. To mitigate these bottlen...
Introducing Beethoven — an open-source full-stack framework that simplifies the development and deployment of hardware accelerators on FPGA and ASIC platforms. Come check out our inaugural Beethoven tutorial at ISCA 2025 in Tokyo, Japan! To learn...
6th-year CS PhD Student
Chris is the lead architect of Beethoven, an open-source accelerator development and deployment framework. His research leverages modern hardware description languages’ flexible, generative ability and programming abstraction to provide scalable and reusable SoC infrastructure for hardware accelerator development.
5th-year CS PhD Student
Mason is the lead architect of PyTFHE, an open-source compilation and execution framework for Fully Homomorphic Encryption (FHE) applications. His research focuses on efficient software and hardware design for privacy-preserving computing, with a particular emphasis on advancing FHE through optimizations in arithmetic, compilers, and hardware accelerators.
4th-year ECE PhD Student
Mansi is the creator of COCOSSim, a cycle-accurate, offload, configurable, and open-source systolic-array-based accelerator simulator. Her research focuses on workload analysis, performance modeling, and hardware acceleration through architectural and microarchitectural enhancements for domain-specific applications, including artificial intelligence.
2nd-year CS PhD Student
Ning's research investigates hardware and software optimizations for more efficient and accurate LLM serving systems. Specifically, his work focuses on accelerating vector databases and retrieval-augmented generations.
Former PhD Student. Now: Postdoc at HKUST
Entropy is the lead architect of ProSE, a protein discovery engine that executes ProteinBERT inference at significantly better power and performance efficiency compared to SOTA GPU and TPU platforms. He is also the creator of fast, accurate, and transferrable synthesis predictors SNS and SNSv2. His research focuses on computer architecture for AI and AI for computer architecture.
