Dotscience

Our vision: DevOps for ML

For AI to be successful it must be reproducible, accountable, collaborative and continuously delivered.

1. Reproducible AI can't be productive or safe without reproducibility. This is hard in ML because it has many more variables.

2. Accountable When models make critical decisions, they must be accountable for how they were made. Models without full provenance can fail compliance regulations.

3. Collaborative For scalable ML teams you must be able to collaborate effectively to avoid confusion, rework and errors.

4. Continuous You are not done when you ship. Models must be retrained and statistically monitored for drift.