research REPACSS Co-designed and built a $12.25M NSF-funded supercomputing facility powered by renewable energy, from rack layout to system software. Secure & Autonomous HPC AI-driven monitoring, anomaly detection, and layered cyber-defense for open-science infrastructure. Energy-Efficient AI Infrastructure Benchmarking and tuning the power consumption of LLM inference and GPU clusters. software MonSTer An out-of-the-box monitoring framework for HPC systems, adopted by Dell's Omnia project. Disaggregation-Aware Scheduler A simulation framework for job scheduling in HPC systems with disaggregated memory, developed with LBNL. xBGAS Simulation Toolchain Cycle-accurate simulator and runtime for the xBGAS global address space extension to RISC-V.