The Multiscale Advanced Training in Research Intensive Computing at Stanford Bootcamp
MATRICS is a week-long, hands-on High Performance Computing (HPC) bootcamp for Stanford graduate students and postdocs, and is a collaboration between the SDSS Center for Computation, Stanford HAI/Data Science, and NVIDIA, with generous computing support from Google Cloud and NSF ACCESS. It is designed to provide you with the real-world skills to take full advantage of Stanford's world-class computing infrastructure through your own research.
MATRICS 2026 is taking place from May 11-15 on Stanford Main Campus. Welcome to the inaugural MATRICS Cohort! We're excited to see you at MATRICS soon!
MATRICS covers the full HPC stack needed for academic level research, from running your first batched job to running large-scale parallelized AI workloads on computing systems like Stanford's Sherlock cluster, NSF ACCESS's cluster network, NVIDIA Cloud, and Google Cloud.
Morning sessions are composed hands-on tutorials where participants apply new HPC concepts to real examples live on supercomputing clusters. Sessions are led by the MATRICS tutorial team, and are structured as guided peer-learning environments. Topics include cluster architecture and job scheduling, parallelization strategies, machine learning workflow optimization, and more.
Afternoon sessions are dedicated project time. Participants bring a project from their own dissertation or postdoctoral research, and apply the morning's skills directly to their own workflows, with facilitation and assistance from the SDSS Center for Computation Team, SRCC, and NVIDIA Solutions Architects.
| Repository | Description |
|---|---|
estimating-and-requesting-resources |
How to estimate the HPC resources that you'll need, and efficient pathways to request them |
stacking-and-moving-containers |
How to move your software stack from machine to machine without reinstallation |
hyperparameter-sweeps-with-wandb |
How to run and track hyperparameter sweeps automatically across HPC systems |
scaling-up-cpus-and-gpus |
How to distribute your code across multiple CPUs and/or GPUs |
preempting-and-checkpointing |
How to run your job on partitions that are "interruptable" |
Repositories are under active development. Stay tuned for more repos!
MATRICS is open to Stanford graduate students and postdoctoral researchers. No prior HPC experience is required and participants should have intermediate to advanced programming skills in any language.
The inaugural MATRICS cohort is already filled. The next round of applications will open in early January 2027. We're excited to see you there! Look out for announcements about the next application season at the beginning of Winter Quarter.
Questions? Reach out to the MATRICS Bootcamp Team at matrics-bootcamp@stanford.edu.
The MATRICS 2026 Bootcamp Team: Ellianna Abrahams, Brian Chivers, Patrick Hynes, Christina Gancayco, Brian Tempero, Zoe Ryan, and Amanda Butler
We gratefully acknowledge the generous support of NVIDIA and Google Cloud to help make this bootcamp possible, and thank the NSF for granting us the computing allocation CIS260896 from the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program, which is supported by U.S. National Science Foundation grants #2138259, #2138286, #2138307, #2137603, and #2138296.
