System Design for Information Retrieval
Details and sub-projects for System Design for Information Retrieval.
System Design for Information Retrieval
HW/SW co-design for Large Scale Retrieval Systems
Research Overview
Hardware-software co-design for large-scale information retrieval and RAG systems is a central direction in the lab. The group explores scalable representations across graph, table, and retrieval-augmented model pipelines, including domain links such as bioinformatics. The effort spans system architecture, indexing, serving optimization, and end-to-end performance modeling so retrieval stacks can scale with better latency-throughput characteristics and stronger deployment reliability.
Research Details
Current efforts in this area include GraFetch: Accelerating Graph Applications Through Domain Specific Hierarchical Hybrid Prefetching: Retrieval- and graph-system acceleration through prefetching and co-design. , Information Retrieval: Recent Advances and Beyond: Systems-centric direction for modern large-scale retrieval. , Accelerating Retrieval-Augmented Generation: End-to-end acceleration of retrieval and generation pipelines. , RAGO: Systematic Performance Optimization for Retrieval-Augmented Generation Serving: Structured optimization for production RAG serving. , and Accelerating Graph Indexing for ANNS on Modern CPUs: CPU-centric indexing and serving optimizations. .