Three Engineering Faculty Members Receive Amazon Research Awards

Jan. 6, 2026 – Amazon selected three UC Irvine electrical engineering and computer science faculty for Research Awards. Sitao Huang, Zhou Li and Yanning Shen are among 63 recipients from 41 universities around the world to win the competitive award that includes unrestricted funds and Amazon Web Services (AWS) promotional credits.

The awardees have access to more than 700 Amazon public datasets and can use AWS artificial intelligence/machine learning services and tools through their credits. They also will be assigned an Amazon research contact who offers consultation and advice, along with opportunities to participate in Amazon events and training sessions.

Huang, an assistant professor, and Shen, an associate professor, will receive 50,000 AWS computing credits for their project, “Automatic Kernel Synthesis and Tuning for AWS Trainium via Profile-Guided Graph Topology Optimization.” With the emergence of novel deep learning acceleration (DLA) chips and systems, the software compilation flow faces a serious challenge from the rapidly increasing diversity of new deep learning models (especially large language models) and the complexity of novel computer architectures. Mapping diverse, large-scale, novel machine learning (ML) models into emerging hardware accelerators has posed significant challenges to the compiler design. Huang and Shen propose to research the next-generation ML compilation flow that automatically optimizes performance of machine learning programs. They plan to model tensor program synthesis and optimization as a graph topology optimization problem, and leverage profile-guided retrieval-augmented generation for domain-specific code generation and optimization. The goal of the project is to elevate the performance of machine learning programs and the programming productivity of software programmers.  

Li, an associate professor, will receive $80,000 and 40,000 AWS cloud credits for his project titled “Precise and Analyst-friendly Attack Provenance on Audit Logs with Large Language Models (LLM).” Li’s goal with this project is to advance the state of attack provenance, an important technique that analyzes the audit logs collected from hosts and reconstructs the attack chains. Despite its capability of detecting sophisticated, nation-state attacks, attack provenance still suffers from low precision when analyzing large-scale, noisy audit logs. The outcomes of the existing works are either alerts and/or their interactions with other system entities, which are far from the actionable threat intelligence needed by human analysts. In this project, Li will explore the integration between attack provenance and LLM, and work to make attack provenance precise and human-friendly.

– Lori Brandt