AI curriculum

Course outline for Cloud Foundations for AI Engineers

A hosted model API and self-hosted inference assign cost, data handling, and recovery to different owners. Design a basic AI application deployment with explicit compute, network, storage, and identity boundaries. Compare service choices at two utilization levels while accounting for token charges, rate limits, and durable application state.

Technology: Cloud and Platform Engineering
Duration: 4 hours
Modules: 2 · 2 hours each
Level: Level 1

About Cloud Foundations for AI Engineers

Cloud Foundations for AI Engineers Course Objectives

Map compute, storage, networking, and identity for an AI application.
Separate durable conversations and source data from ephemeral runtime state.
Compare hosted model APIs, serverless application code, and self-hosted inference responsibilities.
Identify model-service permissions, rate limits, and data-residency assumptions.
Estimate token/API and self-hosted operating costs at two utilization levels.
Define acceptable data loss, recovery time, and a restoration check.

Pre-requisites

Basic application and networking awareness.
Familiarity with operating-system concepts and application data storage.

Before this course

Complete the required courses or bring equivalent experience. Recommended courses add useful depth; they are not enrolment gates.

No prerequisite course is required. The entry skills above still apply.

Has basic application, networking, and OS awareness; this is the cloud entry course.

Find your place in the learning map

Lab Setup

Browser, spreadsheet, and diagram editor.
Supplied hosted-model/self-hosted service descriptions, token and capacity cost inputs, access/data-location scenarios, and rate-limit/restoration records.
Tokens and inference are introduced through a short supplied glossary; no AI prerequisite, cloud account, GPU provisioning, or live model is required.

Detailed Course Outline

Module 1: Map an AI application's infrastructure

Compute: Separate AI application code from inference compute; compare hosted API and self-hosted runtime responsibilities without assuming ephemeral execution preserves conversation state.
Storage: Inventory conversations, source documents, and required backups; distinguish durable application records from temporary inference/runtime files and incomplete restoration evidence.
Networking: Map application-to-model and source-data paths; inspect exposure, transfer, and data-location assumptions without treating a region label as complete compliance evidence.
Identity: Define application and model-service permissions using dummy identities; reject broad credentials and distinguish API reachability from authority to access source data.

Practical outcome: Sketch an AI assistant deployment with model-service access, durable conversation/source records, network boundaries, and data-location assumptions using supplied service descriptions.

Module 2: Compare AI service cost and recovery

Managed services: Compare hosted model APIs with self-operated inference; record provider retention, service limits, and application-owner recovery responsibilities from supplied descriptions.
Serverless: Place application handlers around a model API and compare ephemeral state, execution limits, and utilization-sensitive costs; do not assume arbitrary inference fits serverless constraints.
Cloud economics: Compare input/output token charges with self-hosted capacity at two utilization levels; include idle compute, storage, transfer, and operator effort under stated assumptions.
Reliability: Inspect model-service rate limits and runtime loss; specify bounded retries and durable-state recovery while distinguishing provider redundancy from demonstrated application restoration.

Practical outcome: Compare hosted APIs and self-hosted inference at two utilization levels, inspect a rate-limit fixture, and define application-state recovery targets without assuming provider redundancy restores user data.

Practical exercise

Produce a deployment diagram, durable-state inventory, access matrix, and hosted-versus-self-operated responsibility comparison for a synthetic AI assistant.
Accept intended model-service access; reject an overbroad identity and unintended source-data exposure, and flag an unresolved data-residency requirement.
Calculate cost ranges at two utilization levels using supplied input/output token counts and inference capacity assumptions; include storage, transfer, idle capacity, and operator effort.
Inspect rate-limit and runtime-loss fixtures; define bounded retry behavior, check backup completeness, and separate recovery targets from demonstrated restoration results.

What you will produce

Design an AI application deployment with explicit service ownership, durable state, access/data-location boundaries, cost assumptions, and recovery requirements.

How the work is checked

The diagram and inventory identify every required persistent artifact.
Overbroad access and unintended exposure are rejected.
Two utilization comparisons include token/API costs, self-hosted capacity, and operator responsibilities; rate-limit and data-location assumptions are explicit.
Recovery targets are distinct from observed restoration results and redundancy.

Four-hour architecture and simulation lab teaching cloud fundamentals through an AI assistant. Use supplied token/capacity figures rather than benchmarking inference or configuring GPUs. Data-location checks do not certify compliance. Detailed model serving follows in AI Infrastructure; no live provisioning or production recovery guarantee is included.

When another approach is better

Keep a simple local deployment when cloud operating cost and complexity exceed the workload's needs.
Do not choose an ephemeral runtime for the only copy of essential application data.

How we train

Ask about exercises, instructor feedback and the prior knowledge you need. Tool-specific courses marked provisional may become modules in a broader course.

Discuss your learning goals