Introduction to NAIC WP7 Demonstrators

Objectives

• Understand what Work Package 7 delivers

• Learn about the seven use cases and their scientific domains

• Know where to find each repository and tutorial

• Understand the common infrastructure pattern

Overview

The Norwegian AI Cloud (NAIC) promised that shared infrastructure could lower the barrier for researchers to apply machine learning in their own domains. Work Package 7 (WP7) tested that promise by building seven demonstrator projects, each solving a real research problem and each packaged as a self-contained pipeline that runs on NAIC Orchestrator VMs.

This deliverable — D7.10 — summarizes all seven demonstrators: their scientific contributions, methodology, infrastructure requirements, and current status.

The Seven Use Cases

UC	Title	Domain	Key Technique
UC1	Climate Indices Teleconnection	Climate Science	Ensemble ML (RF, XGBoost, MLP)
UC2	PEM Electrolyzer PINN Optimizer	Green Hydrogen	Physics-Informed Neural Networks
UC3	Pseudo-Hamiltonian Neural Networks	Dynamical Systems	Port-Hamiltonian Decomposition
UC4	3D Medical Image Registration	Medical Imaging	ANTsPy + HD-BET
UC5	Graph-Based AIS Classification	Maritime Surveillance	Graph Neural Networks (DGL)
UC6	Multi-Modal Optimization	Optimization	Hybrid GA + CMA-ES
UC7	Latent PDE Representations	Scientific Computing	Autoencoders + Latent Alignment

Repositories

All repositories are hosted on GitHub under the NAICNO organization:

UC	Repository	Tutorial
UC1	wp7-UC1-climate-indices-teleconnection	Tutorial
UC2	wp7-UC2-pem-electrolyzer-digital-twin	Tutorial
UC3	wp7-UC3-pseudo-hamiltonian-neural-networks	Tutorial
UC4	wp7-UC4-medical-image-registration	Tutorial
UC5	wp7-UC5-ais-classification-gnn	Tutorial
UC6	wp7-UC6-multimodal-optimization	Tutorial
UC7	wp7-UC7-latent-pde-representation	Tutorial

Common Workflow

Every completed demonstrator follows the same operational pattern:

1. Provision a VM on orchestrator.naic.no with GPU support

2. Clone the repository and run the setup script

3. Launch Jupyter via SSH tunnel

4. Run the self-contained notebook end-to-end

5. Inspect results — figures, metrics, and saved models

This pattern means a researcher can go from git clone to results without managing infrastructure or installing complex dependencies manually.

Technology Stack

Technology	Use Cases	Role
Python	All	Primary implementation language
PyTorch	UC2, UC3, UC5	Deep learning framework
TensorFlow	UC7	Deep learning framework
scikit-learn / XGBoost	UC1	Classical ML and gradient boosting
DGL	UC5	Graph neural network library
ANTsPy / HD-BET	UC4	Medical image registration and brain extraction
CMA-ES / DEAP	UC6	Evolutionary optimization

Infrastructure

Infrastructure	Use Cases	Purpose
NAIC Orchestrator VMs	All	GPU-enabled cloud VMs for interactive development
Jupyter Notebooks	All	Interactive demonstrator interfaces
Sphinx Tutorials (GitHub Pages)	All	Multi-chapter tutorial documentation
AI Agent Files	UC1, UC2, UC3, UC5, UC6, UC7	AI coding assistant integration (AGENT.md)

Contributors

Institution	Contributors	Use Cases
NORCE Research	Klaus Johannsen, Odd Helge Otterå, Adrian Evensen, Hasan Asyari Arief, Xue-Cheng Tai, Gro Fonnes, Nadine Goris, Bjørnar Jensen, Jerry Tjiputra, Yngve Heggelund	UC1, UC2, UC5, UC6, UC7
SINTEF Digital	Sølve Eidnes, Kjetil Olsen Lye	UC3
UiB	Saruar Alam	UC4

What You Will Learn

Episode	Topic
02	Provisioning a NAIC VM
03	Getting started with any use case
04–10	Detailed walkthrough of each use case
11	Cross-cutting patterns and lessons learned
12	FAQ

Keypoints

• WP7 delivers seven self-contained ML demonstrators across diverse scientific domains

• All demonstrators run on NAIC Orchestrator VMs with GPU support

• Each repository includes data, code, environment specs, and a Jupyter notebook

• All demonstrators include Sphinx tutorials published on GitHub Pages

• Physics-informed approaches (UC2, UC3, UC7) generalize better with fewer parameters

• UC1 provides both interactive notebooks and CLI parameter sweeps