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.
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
| Artifact | Description | Domain | Status | Links |
|---|---|---|---|---|
| Gemma_3NCORE | Cross-architecture knowledge distillation from MedGemma to Gemma 3n for cardiac imaging | Healthcare/Cardiac Imaging | ✅ Available | Colab |
- Browse the Collection: Explore the artifacts directory for available implementations
- Choose an Artifact: Each artifact contains comprehensive documentation and setup instructions
- Follow Setup Guide: Use the artifact-specific README for detailed instructions
- Apply DEITY Principles: Ensure ethical AI development, especially for healthcare applications
- Python 3.8+
- GPU access (Google Colab A100 recommended for most artifacts)
- Hugging Face account (for model access)
- Appropriate dataset permissions
- 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
We welcome contributions that align with the ATRIOM methodology and DEITY principles. See CONTRIBUTING.md for guidelines.
This repository is for research purposes only. Each artifact may have specific licensing requirements - please check individual artifact folders for details.
This collection is made possible through open-source contributions from the ML community, including datasets, models, and frameworks used across various artifacts.
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.