Robotics labs capped at 6–7 training runs/day on desktop GPUs. Shared campus compute lifted that to 100+ runs daily.
One of the country's leading research universities, with twelve schools simultaneously pursuing federally funded research across biomedical science, robotics, engineering, and the arts and sciences.
Public cloud could address short-term compute gaps but proved too expensive for the iterative experimentation rigorous AI research demands,...
“With the DGX SuperPOD, we were able to improve Vision-Language-Action (VLA) model training speed by 12X, scaling models to 3B+ parameters previously out of reach.”
Private Ivy League research university providing academic and research programs.
Jointly deployed Penn's AI compute, high-performance networking, and supporting infrastructure.
NVIDIA is a technology company that specializes in semiconductors, graphics processing units, and artificial intelligence for applications in data centers, gaming, and more.
University of Pennsylvania's Shared research computing is part of this use case:
Related implementations across industries and use cases
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.
Compute limits capped labs at 1–2 targets/year. A shared 760-GPU cluster let one team screen 10M molecules in a week.
Siloed spreadsheets held back research for years. AI now scans 100M+ data points to guide wound care, cutting analysis cycles to weeks.
Strict privacy rules bottlenecked AI rollouts. Now, a hybrid self-service platform dynamically routes tasks by data sensitivity.
Researchers lost months debugging network failures. Bare-metal clusters ended the crashes and enabled custom orchestration.
Consultative selling couldn't scale. AI now tags thousands of meeting recordings to targeted skills so managers can actively coach.
Agents juggling 8+ tools struggled to keep pace. Now, AI handles 35% of chats, letting the team resolve complex issues 58% faster.
Tournaments running simultaneously meant an hour of manual checks each. AI agents now run them in minutes, freeing the team to be proactive.
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.
Robotics labs capped at 6–7 training runs/day on desktop GPUs. Shared campus compute lifted that to 100+ runs daily.
One of the country's leading research universities, with twelve schools simultaneously pursuing federally funded research across biomedical science, robotics, engineering, and the arts and sciences.
Public cloud could address short-term compute gaps but proved too expensive for the iterative experimentation rigorous AI research demands,...
“With the DGX SuperPOD, we were able to improve Vision-Language-Action (VLA) model training speed by 12X, scaling models to 3B+ parameters previously out of reach.”
Private Ivy League research university providing academic and research programs.
Jointly deployed Penn's AI compute, high-performance networking, and supporting infrastructure.
NVIDIA is a technology company that specializes in semiconductors, graphics processing units, and artificial intelligence for applications in data centers, gaming, and more.
University of Pennsylvania's Shared research computing is part of this use case:
Related implementations across industries and use cases
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.
Compute limits capped labs at 1–2 targets/year. A shared 760-GPU cluster let one team screen 10M molecules in a week.
Siloed spreadsheets held back research for years. AI now scans 100M+ data points to guide wound care, cutting analysis cycles to weeks.
Strict privacy rules bottlenecked AI rollouts. Now, a hybrid self-service platform dynamically routes tasks by data sensitivity.
Researchers lost months debugging network failures. Bare-metal clusters ended the crashes and enabled custom orchestration.
Consultative selling couldn't scale. AI now tags thousands of meeting recordings to targeted skills so managers can actively coach.
Agents juggling 8+ tools struggled to keep pace. Now, AI handles 35% of chats, letting the team resolve complex issues 58% faster.
Tournaments running simultaneously meant an hour of manual checks each. AI agents now run them in minutes, freeing the team to be proactive.
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.