AI Cloud EngineeringMLOps
A model cycle that tells a care team who to call this week, and why
Which two hundred patients do I call this week, and what does it save?
A medical director running a chronic care programme pays thousands of euros for every avoidable readmission and can afford a few hundred preventive nursing calls a week. Today the call list is built from rules of thumb, so the calls reach patients who would have been fine and miss those who come back within a month. This platform ranks the patients to call, prices each call in expected cost avoided, explains every score in plain language and audits who the list reaches. All the data are synthetic.

- precision among the patients contacted
- 0.758
- patients on the weekly call list
- 200
- plain reasons behind every score
- 3
Synthetic data throughout
Gallery
Architecture of the care platform
Built with
Python
Docker
GitHub Actions
FastAPI
scikit-learn
SHAPMLflow
Bicep
Azure Machine Learning
PostgreSQL
Container Apps
Managed identities
Next project
Multi-Tenant AI SaaS Platform
One AI service for many companies, each isolated, metered and capped