Bhushan Chitre

About

Bhushan Chitre is a Research Cyberinfrastructure Architect in Research Computing Infrastructure (RCI) at the University of Kentucky. He designs, deploys, and runs the university’s research computing systems, which serve more than 250 principal investigators across 65+ departments and 11 colleges, plus partner institutions across Kentucky. His work covers HPC clusters, GPU-accelerated platforms, parallel storage, and high-speed networks.

Bhushan has been at UK for more than 8 years. He helps run and tune the university’s research computing environment: the LCC, MCC, ECC, and DGX clusters, with about 5 petaflops of compute, 35,000+ CPU cores, 300+ NVIDIA GPUs, and 55 PB of raw storage. In FY2025 these systems ran nearly 12 million jobs and 111 million compute hours. He holds an M.S. and a B.S. in Computer Science from the University of Kentucky.

Areas of Expertise

• HPC and GPU cluster architecture, deployment, and operations
• AI/ML infrastructure for distributed training and shared model-serving (inference) platforms
• NVIDIA DGX BasePOD (H100/H200) with Base Command Manager; DGX Spark (GB10 Grace Blackwell)
• SLURM scheduling: fairshare, QOS, GPU allocation, partitions, and preemption
• InfiniBand (HDR/NDR) networks and GPUDirect RDMA
• Parallel file systems and storage: IBM Storage Scale (GPFS), Ceph, Dell PowerScale, ZFS, and tape archives
• Automated cluster setup and configuration with Warewulf/OpenHPC and Ansible
• System monitoring with Prometheus, Grafana, and Alertmanager
• Containers for research workloads: Apptainer/Singularity, Docker, Pyxis/enroot, and Kubernetes
• Research data transfer with Globus

Current Projects

• NVIDIA DGX BasePOD, Institute for Biomedical Informatics: an 8-node, 64-GPU H100/H200 cluster on an NDR InfiniBand network with 12 PB of Dell PowerScale storage, supporting biomedical AI workloads such as AlphaFold, Parabricks, and BioNeMo.
• CAPTIVATE Data Bank (NSF EPSCoR E-CORE): infrastructure lead for the storage and archive layer of a statewide climate and environmental research data repository. It combines IBM Storage Scale on Lenovo DSS-G (3.56 PB raw) with a Spectra Logic LTO-10 tape archive (about 2.8 PB usable).
• DGX Spark fleet and AI inference platform: an 8-node NVIDIA GB10 fleet with shared large language model (LLM) services for UK researchers.

Programming Languages and Tools

C/C++, Python, Bash, JavaScript/TypeScript; FastAPI, React, Django; REST APIs, OIDC/OAuth (CILogon); MySQL, PostgreSQL, MongoDB

Education

• M.S., Computer Science, University of Kentucky (2017)
• B.S., Computer Science, Minor in Mathematics, University of Kentucky (2015)

Selected Publications

• Gazula, Lamichhane, Chitre, Quick, Griffioen. “Using AI-coding to Quickly Create Custom System Administration Tools.” PEARC 2026.
• Chappell, Chitre, Gazula, Pike, Griffioen. “Decoupling OpenHPC Critical Services.” HPCSYSPROS Workshop at SC19, 2019.
• Rivera et al. “Navigating the Unexpected Realities of Big Data Transfers in a Cloud-based World.” PEARC 2018.
• Chitre, Hayes, Dekhtyar. “Second-Guessing in Tracing Tasks Considered Harmful?” REFSQ 2018.

Professional Activity

• Supercomputing (SC) conference exhibitor, 2017–2025
• Organizer, Commonwealth Computational Summit, 2017–2024