Doing the work · part of the practice · 3 Hours
A 3-hour hands-on workshop on running ML research at scale — SSH workflows, SLURM job scripts, containerised environments, and choosing between university HPC clusters and cloud GPUs.
Curriculum
3 Hours of live teaching, in the order you'll use it on a real project.
Cluster architecture, login vs. compute nodes, and the SSH workflows every cluster expects.
Requesting GPUs, memory, and walltime, job arrays, and dependency chains for multi-stage pipelines.
Conda/venv, module systems, and Docker/Singularity containers that actually run on shared clusters.
Moving, storing, and versioning large datasets and checkpoints without filling shared storage.
Logging, Weights & Biases, and diagnosing failed jobs without wasting your queue priority.
Spot instances, cost control, and when cloud beats a university cluster.
Data and model parallelism basics, and when you actually need more than one GPU.
Estimating GPU-hours for a project and defending a compute budget in a proposal.
Outcomes
01
Reusable SLURM scripts for single, array, and multi-GPU jobs.
02
A documented SSH/remote-development setup for your own research.
03
A GPU-hour budget for your next experiment or proposal.
04
A checklist for diagnosing failed or stalled cluster jobs.
Tools & frameworks you'll use
Not ready to enrol? Start free
Three job scripts (single GPU, job array, multi-GPU) and a cheat sheet for university HPC clusters.
See what's insideReserve a seat
Pay securely by card and your seat is confirmed instantly — we'll email you the joining details.
Free reading for this course
Keep going
Design experiments and prove your results actually hold up.
You leave with
Third Saturday of every month · live online
OpenAgentic AI pipelines for literature, experiments, and writing.
You leave with
Fourth Saturday of every month · live online
OpenLaTeX, Git, and reproducible code — the toolkit, fast.
You leave with
Third Sunday of every month · live online
Open