Subgraph Matching
-
Using GPUs to accelerate subgraph matching.
-
Using external memory to handle large graphs.
-
Designing sampling algorithms to approximate subgraph matching.
LLM Inference
-
Designing metrics to evaluate the user experience in large language models serving.
-
Developing scheduling strategies to accelerate the inference of large language models.
-
Redesigning the kernel of operators in large language models for extreme performance.
Community Detection
- Using GPUs to accelerate community detection algorithms, such as Louvain and Leiden.
Vector Database
- Using GPUs to accelerate approximate nearest neighbor search in vector databases.