I develop scalable research software accompanying my work in active learning, Bayesian inference, and large-scale density-based clustering.

Approx-FIRAL

Scalable active learning for multinomial logistic regression.

• Matrix-free Hessian operations
• Randomized trace estimation
• Preconditioned conjugate gradients
• Multi-GPU acceleration using MPI and CUDA
• Demonstrated scalability to million-point datasets

Related publication: SC 2024 (Best Student Paper Finalist)
Paper · Zenodo Repository

kNN-DBSCAN

High-dimensional density-based clustering with distributed scalability.

• k-nearest-neighbor graph reformulation of DBSCAN
• Approximate distributed minimum spanning tree
• Hybrid MPI/OpenMP implementation
• Billion-point scalability experiments

Related publication: ACM TOPC 2025
Paper · GitHub Repository