2 of 50

We all do some cool stuff

We all have some problems

Let's connect them

This is an icebreaker

3 of 50

About this talk

Each site briefly introduces itself
Each site presents three cool things they do
… and three problems they want to solve in the future
These become starting points for the rest of the program

Time is very limited now

Interesting ideas from this should become unconference sessions

Advance instructions:

Copy template slides for your organization
Fill in (doesn't have to be perfect, adjust as you'd like)

3 cool things
3 problems

Be prepared to briefly talk (2-4 min)
(Slides serve as further notes)

4 of 50

What happens next

Unconference!

We have a few scheduled talks
The rest of talks need to come from the attendees

Two rooms
4.25 hours to fill (see schedule)

Propose ideas via [TBD]

Unconference example scheduling,

Samwalton9, CC BY-SA 4.0, via Wikimedia Commons

5 of 50

Aalto Scientific Computing

6 of 50

About Aalto SciComp

Currently 18 people

~5 HPC admins
~11 Research Software Engineers
~2 in between

Three main areas

HPC cluster admin
Teaching
RSE

Univ-level infra

School of Science common unit

7 of 50

Aalto C1: Quantum computer

I think it is quite cool that I am setting up networking and backups etc. for a quantum computer here in Aalto! It is part of https://fiqci.fi . - Simppa Äkäslompolo

8 of 50

Aalto C2: SciComp garage

Every workday, we have an online help session 13:00 on Zoom
Usually several people come (doctoral researchers are the biggest group)
It's also a internal team discussion time

Avoids the need to schedule meetings, even though we mostly work online
"foot in door" of other�schools

https://scicomp.aalto.fi/help/garage/

School of Science (orange, left) and doctoral researchers (top) are the biggest support customers

9 of 50

Aalto P2: RSE team projects getting chaotic

Many things going on at once
Things on a researcher schedule, which means "no schedule, can't plan"
If we can't plan, we can't make sure we recover as much cost from 3rd-party funding as promised
Our flexibility and adapting to the schedule of research is one of our benefits

10 of 50

Aalto C3: Open OnDemand

Web-based cluster interface

Pretty common in many clusters

Local transcription service in OOD as an app (speech2text)

Cluster handles login, data storage, etc.

Resource dashboard (past jobs, quota)
Handles data transfer, SSH keys, etc.
There's more possibilities for usability there and available apps can be added

11 of 50

Aalto P1: Compliance and security

Is every piece of research software an IT system? Some people (guess who?) say yes.
If it is, either:

Research slows down enough that we shouldn't bother anymore
Researchers continue ignoring the official rules and buy their own services…

This is taking a huge amount of time for discussion for not much benefit.
Researchers are now forced to conduct self-assessments of research security before even getting the grant, more bureaucracy for no reasons?

12 of 50

Aalto P3: VSCode remote-ssh

VSCode remote-ssh makes a usable user interface

Connection, file transfer, running, editor - all in what people already know

Some issues

Will try to scan every file in the "project" folder - open / and it tries to scan and index whole system
Makes it easy to run on login node and not use the cluster

AI agents

Can run any code on system - not cluster-specific but still…

https://scicomp.aalto.fi/triton/apps/vscode/

13 of 50

Imperial C1: Mini HPC clusters for training

What are the benefits of mini HPC clusters for members of the research community?

How can we build and sustain these clusters?

What software stack(s) work(s) best?

How can we package the software stack to support others building something similar?

AI workloads?! (+AI training?)

DRIFT project miniHPC cluster - part of the VisibleHPC scheme

https://drtp-skills.ac.uk/about-drift/

https://github.com/Visible-HPC/visible-hpc

14 of 50

Aalto Computer Science C1: Jupyter with gpu’s

Teaching for basic students sometimes needs large resources.

=> Existing jupyter over kubernetes hosting tens of courses for 100’s of students

=> lately adding old gpu’s there. For end user (=teacher, students) very simple package with good performance. And less students applying for HPC resources.

15 of 50

About SBL IT at Åbo Akademi

1
Scientific IT support for life sciences
ÅAU researcher, Biocenter Finland -related

17 of 50

About MBC

Core Facility

5 people, senior researchers and project researchers
Bioinformatics consulting
Analysis pipelines for bulk and single-cell RNA-seq data
Part of Turku Bioscience Centre and Biocenter Finland, in collaboration with CSC and Elixir
Base funding from Biocenter Finland through university, project funding from Research Council (FIRI)

Research group

~15 people
Computational tools: R packages etc.
Lot of overlap in activity (and personnel)

18 of 50

MBC C1: 8 Bioconductor packages

…and ~20 other tools

19 of 50

MBC C2: Our own computational infrastructure

HPC cluster

4 local CPU nodes
3 “virtual” nodes on CSC
SLURM
Local and CSC storage

Three stand-alone GPU desktops

20 of 50

MBC P1: Memory problems with R

Single-cell data sets are only getting bigger…

21 of 50

MBC P2: Recruiting

Job titles and descriptions
Eclectic background requirements: HPC, research, biology…

22 of 50

University of Oulu

ICT Continuous Services

23 of 50

About University of Oulu ICT Continuous Services

30 People

2 IT Architects
7 on IT Research support (HPC and related stuff from infrastructure to end user support, Linux workstations and everything else that research group may think, we will try support them).
7 on End user services (Windows and Mac workstations, Azure Virtual Desktop etc.)
5 on Identity management and capacity services (Virtualization platform, storage, linux servers and identity management)
8 on Networking, cloud and M365 services (Windows servers, Azure and campus networks).

Actually 2 HPC Admins to run our cluster (25 nodes).
Organizational position/funding

Mainly funded by ICT budget but we also try to get larger hardware investments from research groups.
We do basically all IT infrastructure at University of Oulu and provide support for research group specific infrastructure.

24 of 50

Oulu C1: Open OnDemand and Open Composer

Easy to use web interface for researchers and students
Linux remote desktop:

High use on coursework and we moved to this solution from x2go

Jupyter notebooks
Open Composer made it easy to preconfigure jobs that end users may run. Like OpenAI Whisper to use to transcribe interviews.

25 of 50

Oulu P1: New working methods and change of mindset

Problems to find out resources to guide researches to use SLURM and batch jobs when possible. Mentality was a long that we will run jobs on interactive sessions on standalone servers. This lead many problems and starvation on resources on single servers.

26 of 50

CSC - IT Center for Science

27 of 50

About CSC

??? - around 700 people at CSC altogether but only a variable fraction work on research computing support
Machine learning specialists, research software engineers, data scientists spread over different groups and units
Funded by ministry funding and research projects (EU, Research Council of Finland etc.)

28 of 50

CSC C1: Roihu is coming!

See the roll up in the corner and ask us more! :)

Also a Zoom session tomorrow Wednesday at 2 pm:

https://csc.fi/en/training-calendar/csc-research-support-coffee-every-wednesday-at-1400-finnish-time-2-2/

29 of 50

CSC C2: Marimo notebooks

We just learned about this, looks cool …

Keeps track of dependencies between cells → solves many problems with Jupyter notebooks
It’s just Python → works nicely with git → you can easily switch to batch jobs in HPC
Nice interactive visualisations

… but we have no experience. Does anyone here use Marimo? What are your experiences?

30 of 50

CSC C3: Bird app: Muuttolintujen kevät

Recent publication: https://doi.org/10.1038/s41559-025-02966-3

31 of 50

CSC P1: Support in collaborative research projects

How to get scientists in research projects as us for more help?

-> How to work together more effectively?

32 of 50

CSC P2: How to get researchers ask questions

We are paid to help even with very basic questions on CSC services, but many researchers are not aware of this or are afraid of asking.

Please tell them to come to us and join the research support sessions on Zoom!

33 of 50

CSC P3: Training definitions

How to manage expectations of participants in training?

How to accurately define the target audience and requirements?

34 of 50

University of Helsinki / UH-RSE

35 of 50

About UH-RSE

3 RSEs
What you do:

Research software development (social science, atmospheric science, genomics)
Consultation (academic + IT-support)
Support (online sessions)

Organizational position/funding

University-wide RSE service
IT-support
Mix of permanent and fixed term positions
LUMI AI Factory

36 of 50

UH-RSE C1: We exist!

Many failed efforts over the years to create RSE group (current approach not ideal…)

37 of 50

UH-RSE C2: Domain knowledge

Managed to hire a mix of expertise

expertise + computing > HPC + expertise

expertise + computing > HPC + computing (knowledge)

38 of 50

UH-RSE C3: Clean slate

Unconventional management structure in IT-support (self-organising)

- also a problem…

39 of 50

UH-RSE P1: Discoverability

Getting the word out about UH-RSE

General discoverability/serendipity in research, even basic research software

40 of 50

UH-RSE P2: Domain knowledge

Domain knowledge does not scale for a “general” RSE service

41 of 50

UH-RSE P3: The other side of the coin

Current frame: Researchers who want to do RSE

Developing frame: RSE who want to do research? (conversions during hiring, UH, LUMI AI Factory and here)

42 of 50

Tampere University

43 of 50

TAU Research data managers

A team of 3 people with full-time permanent positions (+ team lead), all of us have research background with doctorates in different fields

Our team is a part of TAU Research data services, which is part of TAU Library

We support mainly large projects (Horizon Europe, Research Council of Finland) and tailor our support to the needs of the researchers regardless of discipline

We support researchers by writing code (Python, R, C++) and curating data, advising in data processing & sharing platforms, data security and safety (together with other TAU services), open science procedures, etc.

This tailored service is budgeted into the project budgets and is billed from the project on an hourly basis

44 of 50

TAU Cool things

We have had a stable team of 3 research data managers for over 2 years now - it was hard to find people with the skills & interest to be successful in this position long-term
A highly automated MariaDB & Python & Power Automate pipeline to run analyses in CSC ePouta and send aggregated non-sensitive results to a Teams channel with email -> no hassle in producing & managing reports
Leveraging HTML, CSS & JavaScript to present complex results & data summaries to researchers

45 of 50

TAU Challenges

Research data backups - fairly large files (c. 25 GB), 100s of GBs in total to CSC Sensitive Data services (SD Connect). Challenges in transfer time and connections from institution network drives.

46 of 50

Computational Biology Group

Faculty of Medicine and Health Technology (MET)