Industrialize ML from cloud to edge with CI/CD, deployment, and monitoring
This five-day advanced training course enables participants to design, optimize, and deploy robust MLOps solutions in cloud, edge, and serverless environments. Participants will learn how to industrialize ML models, implement CI/CD pipelines, optimize performance, and orchestrate large-scale deployments. The program also covers security, governance, and best practices for monitoring.
Is it for you ?
Data engineers, ML engineers, cloud architects, and advanced data scientists. Contexts: ML industrialization, cloud production, edge AI, and advanced automation.
Prerequisites
Proficiency in Python and ML libraries. Basics of CI/CD. Knowledge of cloud platforms (AWS/Azure/GCP).
What You'll Walk Away With
- ✓ Design advanced MLOps pipelines integrating CI/CD and automation.
- ✓ Optimize ML models and workflows for cloud and hybrid environments.
- ✓ Evaluate and implement serverless and edge AI architectures.
- ✓ Design ML monitoring, observability, and governance strategies.
- ✓ Optimize deployment performance, costs, and scalability.
- ✓ Design a complete, production-ready MLOps project.
Training content
1 Day 1 – Advanced MLOps Architecture
- Complete MLOps architecture: ingestion, training, deployment, monitoring
- Industrial patterns: feature store, versioned models, ML artifacts
- Governance: compliance, auditability
- ML CI/CD pipeline: GitHub Actions/GitLab
- ML packaging: Docker, versioned models
- Training + deployment automation
Lab / Exercise: Setting up a complete ML CI/CD pipeline. Deliverable: functional Git repo.
2 Key points & takeaways:
- Structuring an end-to-end ML pipeline
- Standardizing versioning and artifacts
3 Day 2 – Advanced Cloud Deployment
- Kubernetes for ML: autoscaling, rolling updates
- Cloud pipeline optimization: storage, orchestration
- Secrets and security management
- Containerized model deployment
- Load testing and optimization
- Cloud observability
Lab / Exercise: Deploying a model on a cluster + autoscaling. Deliverable: K8s manifests.
4 Key points & takeaways:
- Knowing how to deploy scalable ML
- Mastering cloud optimization
5 Day 3 – Serverless ML
- Serverless architectures: Lambda/Cloud Run
- Cold starts, limits, optimization
- Serverless feature engineering
- Serverless API inference deployment
- Costs + execution time optimization
- Advanced serverless ML patterns
Lab / Exercise: Serverless inference API. Deliverable: endpoint.
6 Key points & takeaways:
- Minimizing costs + latency
- Integrating ML into serverless workflows
7 Day 4 – Edge AI and Distributed Deployment
- Edge AI: memory and compute constraints
- Model conversion: ONNX, quantization
- Cloud-to-edge synchronization
- Deployment on edge devices
- Device monitoring
- Connection resilience
Lab / Exercise: Quantized model on an edge device. Deliverable: demonstration.
8 Key points & takeaways:
- Adapting ML to constrained environments
- Optimizing compute
9 Day 5 – Monitoring, Governance & Final Project
- Monitoring drift and performance
- Alerts, logs, metrics
- Advanced data governance
- Complete MLOps project
- Deployment + monitoring
- Presentation + technical audit
Lab / Exercise: Complete MLOps project. Deliverable: repo + demo.
10 Key points & takeaways:
- Knowing how to industrialize ML from end to end
- Mastering monitoring + governance
📌 Practical information
Our training sessions are offered in Montreal or Quebec City, in person or in a virtual classroom. Dates and locations are specified when you select your session below. If you have any questions, check out our FAQ.
