MLOps

Dataiku offers numerous capabilities for implementing a complete MLOps practice:

• The Flow offers full lineage and traceability on the design on models

• Deploying projects to production with full CI/CD capabilities

• Visual Machine Learning

• Real-time REST API scoring server with versioning

• Exporting models to Python, Java, MLflow and PMML for scoring outside of DSS

• Comparing models, models versions and behavior of models over time

• Performing advanced drift analysis

• Test scenarios provide the ability to test parts of a project flow, webapps and run Python unit tests.

This section focuses on the following MLOps-specific capabilities:

• Curating features in a Feature Store

• Evaluating and comparing models and model versions

• Analyzing drift

• Importing models from external Machine Learning systems

• Tracking code experiments

• Monitoring projects, endpoints and models health

Note

To get started with MLOps, you may wish to first try the MLOps Quick Start and other resources on MLOps and operationalization in the Knowledge Base.

• Feature Store
• Models evaluations
• Model Comparisons
• Drift analysis
• MLflow Models
• External Models
• Experiment Tracking
• Unified Monitoring