Purdue University · NSF · Intel · DOE & DOD partners
HPC & AI Systems
I co-designed and optimized computing systems from petascale simulations to exascale.
My work spanned scientific workloads, runtime and system software, hardware, data centers, and government partners (DOE and DOD). Together, this work helped shape systems that pushed the limits of computation.
Scaling scientific simulations · Purdue University · 100K+ nodes · 100 TFLOPS · NSF.
Performance · Power · Reliability · Intel · DOE and DOD partners.
A software-and-hardware foundation for building today’s AI products and infrastructure.
How I shaped system outcomes
At scale, performance, power, and reliability have to be designed as one system.
Scaling scientific simulations
At Purdue University, I built petascale HPC software for NSF-supported jet-noise simulation, led a five-researcher team to a 100× performance improvement through a broad portfolio of algorithmic and architectural optimizations—including scalable sparse solvers, point-to-point communication, cache optimization, and parallel I/O. I scaled workloads to 100 TFLOPS across more than 100,000 physical nodes using TACC Ranger (NSF Track-2), NCSA Blue Waters (NSF Tier-1), and Purdue Rossmann. My research advisors were Gregory A. Blaisdell and Anastasios Lyrintzis.
Co-designing exascale systems
At Intel, I led six engineers across workload characterization, performance modeling, power analysis, architecture simulation, runtime design, and resilience through longer-horizon R&D programs including DOE FastForward 1 and 2, DesignForward, CORAL, and DOD PUMA. Separately, we applied that full-stack work to near-term products, including Stampede2 with Intel Xeon Phi KNL and Aurora with Intel Ponte Vecchio GPUs.
From compute to products
That systems perspective became the foundation for my work in consumer AI, voice agents, and agentic infrastructure: find the real bottleneck, align the full stack, and make complex technology useful.