mle-workflow - Production machine-learning engineering workflow
Guides production ML systems through data contracts, reproducible training, evaluation, deployment, monitoring, and rollback.
Tags
Updated: 2026-10-16Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Define data contracts
- Build reproducible training
- Evaluate model quality
- Package deployable artifacts
- Design inference pipelines
- Plan model rollouts
- Monitor production models
- Define rollback paths
- Analyze model errors
- Track data drift
Inputs
- ML project requirements
- Training data
- Feature definitions
- Label definitions
- Model artifacts
- Evaluation metrics
- Serving constraints
- Production traffic
- Monitoring ownership
Outputs
- Data contracts
- Training pipelines
- Evaluation reports
- Versioned artifact bundles
- Prediction endpoints
- Batch scoring jobs
- Rollout plans
- Monitoring dashboards
- Observation ledgers
- Rollback plans
Requirements
- Read, write, edit, shell, search, and file discovery tools
- Project-specific model and data environment
- Access to relevant datasets and artifacts