Cambridge Computer
Cambridge Computer

AI Futures at Harvard Medical School

GPU in EDU at HMS
Event Details

Cambridge and NVIDIA had a full house at Harvard Med for the seminar entitled AI Futures at Harvard Medical School. GPU in EDU is a seminar series co-produced by NVIDIA and Cambridge Computer. We travel to universities and research institutions across the country to share ideas and insights related to AI and ML at the intersection of science and IT infrastructure. The content of each session is tailored to the specific interests of the institution hosting the event. Harvard Med chose to focus on deep learning for genomics, NVIDIA software frameworks, and GPU resource management.

We’d like to thank our friends at Harvard, Run:ai, and NVIDIA for making this event possible. Their support and collaboration were instrumental in making it a memorable and informative event. Special thanks for Run:ai for co-sponsoring the event and sharing insights into their amazing technology for pooling and scheduling GPU resources. We’d also like to thank Christos Alexiadis (Cambridge) and Eliot Eshelman (NVIDIA) for all their work behind the scenes.

Agenda

Welcome Session (10:00am – 10:15 am) Presented by: Eliot Eshelman (NVIDIA) & Jose Alvarez (Cambridge)

Part 1. Deep Learning & Genomics (10:15 am – 12:30 pm) 10:15 – 11:00 am – Training, Tuning, & Deploying large deep learning models (45 mins) – Huiwen Ju (NVIDIA) 11:00 – 11:45 am – Deep learning applications in proteomics (45 mins) – Glen Otero (Cambridge) 11:45 am – 12:30 pm – GPU-based acceleration of genomic data. (45 mins) – Glen Otero (Cambridge)

Lunch break (12:30 pm – 1:30 pm) We will be ordering Milk Street Cafe for our networking/lunch hour portion of the agenda.

Part 2: New NVIDIA Software Frameworks for Life Sciences (1:30 pm – 2:30 pm)

  • Survey of NVIDIA software frameworks in healthcare & life sciences (1 hour) – Huiwen Ju (NVIDIA)
    • NIMs for standardized, efficient Generative AI deployments (hands-on with https://ai.nvidia.com/)
    • BioNeMo for AI-driven drug discovery
    • MONAI the open-source framework for AI in medical imaging
    • VISTA-2D AI Foundation model for cell segmentation
    • RAPIDS for accelerated data science
    • FLARE for federated learning

Coffee Break (2:30 pm – 2:45 pm)

Part 3. GPU Utilization (2:45 pm – 3:45 pm)

  • Summary of latest NVIDIA architectures (5 minutes) – Brad Palmer (NVIDIA)
    • H200, B200, GH200, GB200
  • GPU Monitoring and Management (25 minutes) – Brad Palmer (NVIDIA)
    • Collecting and Displaying GPU performance data
    • Managing GPU configurations across a cluster
  • GPU Resource Management (30 mins) – Brett Wagner (Run:ai)
How to get there