mlops-engineer
Build Production ML Pipelines
Building and maintaining ML pipelines requires expertise in multiple tools and cloud platforms. This skill provides expert guidance on implementing end-to-end MLOps workflows with experiment tracking, model versioning, and automated deployment.
Baixar o ZIP da skill
Upload no Claude
Vá em Configurações → Capacidades → Skills → Upload skill
Ative e comece a usar
Testar
A utilizar "mlops-engineer". Design a basic ML pipeline using Kubeflow
Resultado esperado:
- Pipeline Structure: Data Ingestion → Preprocessing → Training → Evaluation → Model Registration
- Each component runs as a Docker container with appropriate resource limits
- MLflow integrated for tracking metrics and artifacts across all stages
- Pipeline parameters defined for data paths, model hyperparameters, and thresholds
A utilizar "mlops-engineer". How do I set up model versioning with MLflow?
Resultado esperado:
- 1. Register model in MLflow Model Registry with unique version
- 2. Add model metadata including description and training parameters
- 3. Configure stage transitions: None → Staging → Production
- 4. Implement approval workflow for production promotions
- 5. Set up webhooks for notifications on model updates
Auditoria de Segurança
SeguroPrompt-only skill with no executable code. Static analysis scanned 0 files and detected 0 security issues. The skill provides MLOps guidance through text-based instructions only. No network requests, file system access, or external commands. Risk score: 0/100.
Pontuação de qualidade
O Que VocĂŞ Pode Construir
Design ML Platform Architecture
Create a comprehensive MLOps platform design for an enterprise needing to deploy models at scale with experiment tracking and model versioning.
Implement Automated Retraining Pipeline
Build an automated pipeline that retrains models when data drift is detected, with approval workflows and rollback capabilities.
Configure Multi-Cloud ML Infrastructure
Set up ML infrastructure across AWS, Azure, and GCP with consistent tooling and disaster recovery capabilities.
Tente Estes Prompts
Design a basic ML pipeline using Kubeflow that includes data preprocessing, model training, and model evaluation stages. Include configuration for experiment tracking with MLflow.
Set up MLflow experiment tracking for a PyTorch training project. Include configuration for metrics logging, artifact storage, and model versioning.
Design a production ML deployment architecture on AWS SageMaker with auto-scaling, monitoring, and blue-green deployment capabilities. Include cost optimization strategies.
Design a complete MLOps platform architecture including: feature store, experiment tracking,/CD pipeline, monitoring model registry, CI, and drift detection. Specify tools and integration points for AWS or GCP.
Melhores Práticas
- Start with experiment tracking before building complex pipelines to understand data and model behavior
- Implement monitoring and alerting from the initial deployment, not as an afterthought
- Use infrastructure as code (Terraform, CloudFormation) for reproducible ML environments
Evitar
- Deploying models without automated testing or validation gates
- Skipping data quality checks before model training leads to poor model performance
- Using single cloud provider without considering vendor lock-in for critical ML systems