Research

My research is interdisciplinary, spanning high-performance computing, graph and sparse algorithms, and large-scale machine learning, with applications across scientific computing, data science, and AI systems. At a high level, my group focuses on designing scalable algorithms and systems that make emerging data- and compute-intensive workloads practical on modern supercomputers. This work integrates ideas from numerical linear algebra, graph theory, distributed systems, and machine learning, and is motivated by real-world applications in science and engineering. For a complete and up-to-date list of publications, please see my
👉 Google Scholar profile.

1. Sparse Linear Algebra at Scale

We focus on developing high-performance sparse matrix kernels and algorithms for CPUs, GPUs, and distributed-memory systems. For example, we develop scalable SpMM and SpGEMM algorithms and the fusion of various sparse matrix operations needed by graph algorithms, scientific computing, and machine learning.

Representative Publications

• Filler Paper Title on Scalable SpGEMM, Conference YYYY

2. Graph Algorithms and Network Analysis

We develop scalable graph algorithms for clustering, partitioning, and analysis of massive networks.

Representative Papers

• Filler Paper Title on Distributed Graph Clustering, Conference YYYY

3. Scalable Knowledge Graphs and Graph RAG

We develop scalable algorithms and systems for training and inference of large-scale knowledge graphs and GraphRAG models.

Representative Papers

• Filler Paper Title on Distributed ML Training, Conference YYYY
• Filler Paper Title on Communication-Efficient ML, Journal YYYY

4. Interpretable and Explainable Graph ML

::contentReference[oaicite:3]{index=3}

Focus

Developing scalable and interpretable methods for understanding predictions made by graph neural networks and related models.

Sub-projects

• Distributed explainability algorithms
• Shapley-value-based explanations for graphs
• Scalable GNN interpretation pipelines
• Trade-offs between accuracy, cost, and interpretability

Representative Papers

• Filler Paper Title on GNN Explainability, Conference YYYY
• Filler Paper Title on Scalable Model Interpretation, Journal YYYY

5. HPC for Scientific and Data-Driven Applications

We apply HPC and scalable algorithms to real-world scientific and engineering problems including bioinformatics, earth science, and energy science.

Representative Papers

• Filler Paper Title on HPC Scientific Applications, Journal YYYY
• Filler Paper Title on Data-Driven Science, Conference YYYY