HPC and Data Science Summer Institute 2025

• Preparation Day (virtual): Tuesday, July 29, from 9 am to 11 am (Pacific Time)
• Summer Institute (in-person):
  • Monday, August 4 – Thursday, August 7, from 8 am to 5 pm (Pacific Time) 
  • Friday, August 8, from 8 am to 12:30 pm  (Pacific Time)
• Location: San Diego Supercomputer Center, University of California San Diego campus

Overview

The SDSC Summer Institute is a week-long workshop hosted annually by the San Diego Supercomputer Center (SDSC) at UC San Diego, focusing on introductory-to-intermediate topics in High Performance Computing (HPC), Data Science, and Artificial Intelligence (AI). Designed for researchers and educators in academia and industry, the program equips participants to solve scientific challenges exceeding local computing capabilities.

The purpose of the Summer Institute is to give the attendees an overview of topics in High Performance Computing and Data Science and accelerate their learning process through highly interactive classes with hands-on tutorials on the Expanse Supercomputer.  Moreover, the attendees will have many opportunities to meet one-on-one with SDSC’s experts to discuss in detail the best techniques to solve their specific scientific problems. In order to benefit from the classes, the attendees are required to have familiarity with the UNIX/Linux shell. Basic programming skills (in any programming language) are strongly recommended.

Program highlights

See the agenda for more details.

Preparation Day (Virtual – July 29, 2025)

• Orientation: Logging into the Expanse Supercomputer, Linux/Unix basics, and running supercomputing jobs.

Summer institute:

• Monday & Tuesday Morning - Foundational Skills

  • Data Management: File systems, compression, checksums, and secure transfer tools (wget, curl).

  • Batch Job Workflows: Writing Slurm scripts, job scheduling, and troubleshooting submissions.
  • Parallel Computing Basics: Scalability principles, processes vs threads, resource allocation, and benchmarking.
  • High-Throughput Computing: Strategies for managing a large number of small jobs for example for large-scale parameter sweeps.

• Tuesday afternoon to Friday morning - Advanced techniques:

  • Parallel Computing using MPI & Open MP: Learn how to leverage all cores in a machine with OpenMP and how to scale your computations across multiple nodes with MPI using C and FORTRAN.
  • Performance Tuning: Cache optimization, loop-level parallelization, and compiler limitations.
  • GPU Programming: Understand GPU architecture, learn how to program GPUs with libraries, OpenACC directives and CUDA.
  • Deep Learning: Covers core neural network concepts and hands-on Keras implementation, progressing to advanced architectures (e.g., deep CNNs, transfer learning) and GPU-accelerated model optimization, with applications in image/speech recognition and biomedical domains.
  • Python for HPC: Easily speed-up Python on a single machine with numba and then scale a fully distributed workload on a cluster with dask.