Advancing scientific data compression through high-performance computing and machine learning

2D lid-driven cavity streamline visualization - incompressible Navier-Stokes solver

I'm an engineering student working at the intersection of high-performance computing, data compression, and machine learning. Currently, I'm focused on making cutting-edge compression techniques accessible to the scientific computing community through practical, performant implementations.

Undergraduate Research Assistant under Dr. Ranka, working on integrating advanced compression methods into production HPC workflows.

Converting CAESAR, a foundation model for scientific data compression, from Python to C++ for production HPC environments. This work involves:

• C++ Port: Translating the CAESAR compression model to high-performance C++ while maintaining accuracy and improving runtime efficiency
• ADIOS2 Integration: Adding CAESAR as a native compression operator in ADIOS2, enabling seamless compression in large-scale scientific simulations
• HPC Optimization: Designing for scalability and performance in distributed computing environments

Tech Stack: C++, Python, ADIOS2, MPI

Impact: Bringing state-of-the-art learned compression to production scientific workflows, enabling researchers to store and transfer massive simulation datasets more efficiently.

High-Performance Computing

• Parallel computing with MPI
• Data I/O optimization with ADIOS2
• Performance profiling and optimization

Data Compression & Analysis

• Scientific data compression (MGARD, CAESAR)
• Compression accuracy vs. ratio tradeoffs
• Foundation models for compression

Numerical Computing

• PDE solvers (Navier-Stokes equations)
• Computational fluid dynamics
• Scientific visualization with ParaView

• Learned compression for scientific data
• Scalable I/O systems for HPC
• Fluid dynamics simulation and visualization
• Energy-efficient scientific computing
• Bridging ML and traditional HPC