Maintainer of KvikIO, an open-source, high-performance I/O library powering the RAPIDS/cuDF ecosystem. Design and optimize Linux I/O paths. for GPU data ingestion at scale

• Worked on DoorDash Dot at Doordash Labs. Developed a lightweight API to trace GPU activities. Implemented a feature to analyze GPU memory pool fragmentation.

• On the robotics infrastructure team, developed a GPU utility library to facilitate CUDA development.

• For the autonomy AI tasks, developed and optimized CUDA kernels, and performed runtime GPU optimizations.

• Verified segmentation algorithms for the Advance Tumor Imaging Quantification software for FDA review.

• Developed virtual phantom generation workflow.

• Developed a HIP-based library to accurately calculate effective atomic numbers in dual energy CT imaging for colon polyp detection on Nvidia/AMD GPUs.

• Developed an Unreal Engine-based software ElbowCare to optimize baseball athletes’ elbow surgery procedure.

• Maintained a research software 3DQI designed for quantitative medical image analysis. Developed a Python/R-based pipeline to enable routine radiomics analysis and machine learning on medical images for tumor detection and diagnosis.