Nasreddine Hanafi

Research Assistant

ORACLE•  February 2024 - Present

Conducted a comprehensive review of HNSW (Hierarchical Navigable Small World) from Oracle and CAGRA (Cuda-Accelerated Approximate Nearest Neighbor Search Graph-based) from Nvidia. Curated high-dimensional vector datasets for benchmarking. Implemented and benchmarked HNSW and CAGRA, leveraging GPU acceleration for CAGRA and analyzing performance results. Designed and implemented a hybrid indexing strategy, building the index in CAGRA on the GPU and performing the search on the CPU. Documented processes and results, created visual aids, and presented findings to the team.