Design, secure, and scale end-to-end cloud AI applications
This two-day intermediate training course explores the architectures, services, and best practices for designing and operating AI solutions across major cloud platforms (Azure, AWS, GCP). Participants will learn to analyze use cases, implement data and inference pipelines, and configure governance, security, and monitoring. Guided hands-on labs enable the integration of managed AI services, the automation of deployment, and the industrialization of production environments.
Is it for you ?
Data/AI engineers, cloud developers, and technical architects in an enterprise context. Project teams looking to industrialize AI use cases on Azure/AWS/GCP within a governed IT environment. Expected experience: basic programming and cloud concepts; familiarity with ML/AI principles.
Prerequisites
Basic knowledge of Python or equivalent. Understanding of cloud architecture (networking, storage, compute). Familiarity with fundamental ML concepts (training sets, metrics, overfitting).
What You'll Walk Away With
- ✓ Analyze AI use cases and choose the appropriate cloud services (compute, data, managed AI)
- ✓ Implement an end-to-end inference pipeline (ingestion, preprocessing, model, exposure)
- ✓ Configure security, data governance, and costs for AI workloads
- ✓ Integrate managed AI services (vision, language, search) into an application
- ✓ Automate deployment and monitoring (CI/CD, MLOps, surveillance and logs)
Training content
1 Day 1 – Fundamentals of AI in the Cloud and End-to-End Architecture
- Overview of cloud AI services: compute (CPU/GPU/Serverless), data (lakehouse), managed AI
- Reference architecture for an inference pipeline: data flow, preprocessing, model, API, scaling
- Technological choices and criteria: latency, costs, compliance, data residency
- Security and governance: identities, keys/secrets, network segmentation, sensitive data management
- Observability and costs: logs, metrics, traces, sizing and optimization
- Operating model: roles, runbooks, SLA/SLOs, and responsibilities
2 Lab / Exercise:
Design and deploy a minimal inference pipeline in a specific cloud environment: creating necessary resources, deploying a pre-trained model via a managed service, exposing an API, and end-to-end testing. Deliverable: architecture diagram, endpoint URL, test sets, and measured latency/estimated cost.
3 Key points & takeaways:
- Typical architecture of a cloud inference service and its components
- Initial best practices for security, observability, and cost control
4 Day 2 – Integration of Managed AI Services and Industrialization (MLOps)
- Integrating managed AI services (vision, NLP, vector search, embeddings) into an application
- Architectural patterns for RAG and semantic search; managing prompts and contexts
- Securing AI endpoints and protection against abuse (quotas, rate limiting, filtering)
- CI/CD and MLOps automation: packaging, model registries, controlled deployments, rollback
- Monitoring drift, data quality, alerting, and continuous improvement loops
- Ongoing case study: service selection, cost/performance/compliance trade-offs, and operations plan
5 Lab / Exercise:
Integrate a managed AI service (e.g., NLP or vision) into an existing API and set up a simple CI/CD pipeline with basic monitoring (logs/metrics) and functional tests. Deliverable: code repository, operational CI/CD pipeline, monitoring dashboard, and runbook.
6 Key points & takeaways:
- Secure and efficient integration of managed AI services
- First milestones in industrialization with CI/CD and MLOps
- Final assessment of skills: short case study presentation and skills checklist
📌 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.
