GPU-Accelerated Graph Search (HNSW)

*HPC Lab (Tu Lab)

C++, CUDA, Parallel Computing*

Problem: Nearest neighbor search in large-scale biological databases (e.g., DNA alignment) is computationally expensive.
Method: Designed a parallelized search algorithm for Hierarchical Navigable Small World (HNSW) graphs. Optimized memory access patterns and thread divergence on NVIDIA GPUs.
Outcome: Achieved significant speedup over CPU-only implementations.