Rajeev Jain
Rajeev Jain
Research software engineer building high-performance tools for climate science, cancer research, and exascale computing. Lead developer of UXarray (205+ GitHub stars), 2× R&D 100 Award winner, with 16+ years optimizing scientific workflows at Argonne National Laboratory.
Principal Specialist · Argonne MCS Division · Joint appointment at University of Chicago
Featured Work
UXarray: Python library for unstructured climate grid analysis
Challenge: Climate scientists working with unstructured grids (MPAS, ICON, SAM) lacked Python tools for conservative analysis that preserve integral quantities across non-uniform meshes.
Contribution: Lead developer since project inception, implementing core mathematical operators including conservative zonal averaging using Gauss-Legendre quadrature, Grid I/O readers for multiple formats (ESMF, MPAS, SCRIP, HEALPix), and testing infrastructure. Established continuous integration and regular PyPI releases.
Impact: UXarray is now used by researchers at NCAR, DOE labs, and universities worldwide (205+ GitHub stars). Enables analysis of multi-petabyte climate datasets with validated accuracy. Presented tutorials at SC24, AMS 2024, and ESDS Annual Event.
FLASH-X: I/O optimization for exascale multiphysics simulations
Challenge: Checkpoint and restart operations were taking 30-50% of total runtime in billion-element FLASH-X simulations on leadership-class supercomputers.
Contribution: Implemented asynchronous HDF5 I/O with Argobots for non-blocking checkpoint operations and integrated SZ3/ZFP compression. Built verification workflow with nightly baseline testing to ensure reproducibility. Enabled cross-checkpoint restart between AMReX and Paramesh solvers.
Impact: Achieved 40-70% reduction in checkpoint write times on Summit supercomputer. Compression reduced storage requirements by 50%+ with minimal accuracy loss. Published at SC24 workshop, contributed to R&D 100 Award (2022).
FLASH-X Project · FLASH-X Paper · Compression Paper · GitHub
CANDLE/IMPROVE: hyperparameter optimization for cancer drug response models
Challenge: Cancer drug response prediction models showed poor generalization across different pharmacogenomic datasets, requiring systematic benchmarking and optimization.
Contribution: Built hyperparameter optimization (HPO) infrastructure and ran 10,000+ training experiments across Summit, Theta, and Cori supercomputers. Developed GitHub Actions workflows for cross-study validation. Maintained benchmarking framework and co-authored standardization guidelines.
Impact: Benchmarking framework used by 15+ researchers across the project. Results published in Briefings in Bioinformatics (2025) and presented at 20th Workflows Workshop (2025). Contributed to R&D 100 Award (2023).
Recognition
Background
Argonne National Laboratory (2009–present) — Principal Specialist, Research Software Engineering. Lead developer for UXarray, FLASH-X, CANDLE/IMPROVE, MeshKit, and urban simulation projects.
University of Chicago (2023–present) — Staff At-Large. Joint appointment supporting cancer and earth science research.
Education: M.S. Computer Science (University of Chicago, 2020) · M.S. Structural Engineering (Arizona State University, 2009) · Arizona State University Graduate Fellowship (2007-2009)