I was invited to present at the LUMI AI Factory breakfast get-together with PhD students, giving an overview of scientific computing skills opportunities available at Aalto University, in Finland, in the Nordics, and beyond.
The talk started with a reminder about impostor syndrome and the Dunning–Kruger effect: we are all at different points in our life-long learning path, and that is perfectly normal. The goal is not to be the best in everything at once, but to know what exists and where to get help.
The Scientific Computing Skills framework
The talk presented the six-area framework from Hands-on Scientific Computing:
- A. Basics — research workflows, where to get help, setting up your computer
- B. Related science skills — organising research data, Jupyter notebooks, publication-quality figures, LaTeX, scientific posters
- C. Linux and shell — command line, text editors, Git, SSH, Make/SnakeMake
- D. Clusters & HPC — what HPC is, environment modules, Slurm job scheduling, storage, parallel jobs
- E. Scientific coding — modular code, testing, profiling, debugging, software licensing, reproducibility
- F. Advanced HPC — parallel programming models, GPU programming, MPI
Where do we start our scientific computing learning path?
Several organisations offer training and support:
- Aalto Scientific Computing (ASC) — local HPC support and training at Aalto
- CodeRefinery — Nordic network for computational reproducibility
- CSC — Finnish national HPC centre, with a broad training calendar
- Aalto RDM & Open Science Training — open to the world at aalto.fi, recordings on YouTube
- Nordic RSE — Nordic Research Software Engineers community
- LUMI AI Factory — training events including Practical Deep Learning and Moving your AI training jobs to LUMI, see lumi-ai-factory.eu/trainings/
And you can get ECTS!
All of this feeds into SCI-L1010 Scientific Computing Skills, a 5 ECTS umbrella course in Sisu/MyCourses. Aalto University PhD students can enrol, attend the modules most relevant to their work, and earn up to 5 credits within one academic yeaar, using a “pick your own adventure” approach.
The talk closed with a reminder that good enough practices are often better than the best practices. The continuum goes from chaos (a single folder for all projects, no backups, no version control) to best practices (open data with DOI, Git, containerisation, automation). Aiming for “good enough” is a realistic and valuable goal. :)
The slides are available at this link (PDF, 1.7MB).