Skip to main content

Model Deployment

Deploy trained ML model to production including containerization, API endpoint creation, load balancing, versioning, and rollback capability.

Complexity: Complex | Duration: 30m+ | Category: Devops

Tags: ml deployment mlops production

Workflow Diagram​

Steps​

Step 1: Model serialization​

Agent: ml

engineer - Pickle/joblib/ONNX export with version tag

Step 2: Containerization​

Agent: devops

engineer - Create Docker image with model and dependencies

Step 3: API creation​

Agent: ml

engineer - FastAPI/Flask endpoint with input validation

Step 4: Load testing​

Agent: testing

specialist - Stress test with expected QPS

Step 5: Security review​

Agent: security

specialist - Check for vulnerabilities, secrets exposure

Step 6: Deployment​

Agent: devops

engineer - Deploy to Kubernetes/GCP/AWS with blue-green strategy

Step 7: Smoke testing​

Agent: testing

specialist - Verify deployment with sample requests

Step 8: Rollback plan​

Agent: devops

engineer - Document rollback procedure and triggers

Usage​

To execute this workflow:

/workflow devops/model-deployment.workflow

See other workflows in this category for related automation patterns.