ATRIOM Collection

Artifacts Transformation & Resources Interoperability in Machine Learning

A curated collection of proof-of-concept, pilot, and prototype implementations demonstrating MLOps innovations in healthcare applications and beyond.

🫀 Overview

ATRIOM (Artifact Transformation & Resources Interoperability in Machine Learning) is inspired by the heart's atrial functions (ATRIUM + IO):

• Reservoir Phase: Gathering and storing ML artifacts with automated environment setup
• Conduit Phase: Building data pipelines and transforming resources for interoperability
• Active Phase: Training, analyzing, and generating resource-efficient outputs

Each artifact follows the DEITY Principles Framework:

• Data: Transparent, diverse inputs
• Ethics: Responsible AI governance
• Informatics: Interpretable outputs
• Technology: Adaptive solutions
• You: Empowering both human and machine intelligence

📚 Collection

Artifact	Description	Domain	Status	Links
Gemma_3NCORE	Cross-architecture knowledge distillation from MedGemma to Gemma 3n for cardiac imaging	Healthcare/Cardiac Imaging	✅ Available	Colab

🚀 Getting Started

1. Browse the Collection: Explore the artifacts directory for available implementations
2. Choose an Artifact: Each artifact contains comprehensive documentation and setup instructions
3. Follow Setup Guide: Use the artifact-specific README for detailed instructions
4. Apply DEITY Principles: Ensure ethical AI development, especially for healthcare applications

Prerequisites

• Python 3.8+
• GPU access (Google Colab A100 recommended for most artifacts)
• Hugging Face account (for model access)
• Appropriate dataset permissions

📖 Documentation

• ATRIOM Framework - Detailed methodology and phases
• Healthcare Ethics - DEITY principles and ethical guidelines
• Contributing - How to add new artifacts to the collection
• Getting Started - Quick start guide for new users

🤝 Contributing

We welcome contributions that align with the ATRIOM methodology and DEITY principles. See CONTRIBUTING.md for guidelines.

📄 License

This repository is for research purposes only. Each artifact may have specific licensing requirements - please check individual artifact folders for details.

🙏 Acknowledgments

This collection is made possible through open-source contributions from the ML community, including datasets, models, and frameworks used across various artifacts.

👤 Author

Shehab Anwer | The Adimension

• Email: shehab.anwer@gmail.com | shehab@anwer.ch
• X: @ShehabAnwer
• LinkedIn: Shehab Anwer MD
• GitHub: @HABANWER | @The-Adimension

---

© 2025 The Adimension. All rights reserved.