AI curriculum

Course outline for Deployment Foundations for AI Web Applications

An AI web preview can pass while the release changes its model configuration or exposes provider credentials. Compare static hosting with server-side AI backends and track application, prompt, and model-configuration versions together. Test exact-release approval, recovery, and exposure cleanup in a local simulator without creating paid resources.

Technology: AI-Native Web and Application Development
Duration: 4 hours
Modules: 2 · 2 hours each
Level: Level 2

About Deployment Foundations for AI Web Applications

Deployment Foundations for AI Web Applications Course Objectives

Compare static hosting with server-side model calls, streaming, and background-job needs.
Produce an identifiable build with versioned prompt/model configuration and server-side secret references.
Gate promotion on supplied contract/evaluation checks and exact-release approval.
Test repeated promotion and recovery without assuming rollback reverses data or tool effects.
Specify preview/tunnel access, expiry, and cleanup without exposing credentials or sensitive data.

Pre-requisites

Web-feature and packaging competence equivalent to Web Application Foundations for AI Engineers.
Basic Git and command-line build-tool experience.
Familiarity with HTTP and deployment concepts; a full cloud course is helpful but not required.

Before this course

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

Required preparation

Web Application Foundations for AI Engineers

Recommended preparation

Cloud Foundations for AI Engineers

Can build a web artifact and use Git/build tools before designing preview, release, and exposure cleanup.

Find your place in the learning map

Lab Setup

Git, browser, local build tools, supplied mock AI web application, and local release/evaluation-fixture simulator.
Dated hosting profiles, prompt/model configuration fixtures, dummy secret references, and guided preview/recovery cases; no live model, hosting account, or paid plan is needed.
Public tunnels and live deployment require separate authorization, current official instructions, and a cost/access/cleanup plan.

Detailed Course Outline

Module 1: AI hosting fit and release identity

Vercel: Compare dated profiles for AI web handlers, streaming, execution limits, and secret handling; distinguish hosting capabilities from model-provider quotas and data policies.
GitHub Pages: Separate a static AI frontend from server-side model invocation and tool mutations; never embed provider credentials in a published browser artifact.
Cloudflare: Compare Workers-based AI backend and frontend hosting profiles; distinguish application deployment from Workers AI inference, AI Gateway controls, and separately owned data stores.

Practical outcome: Compare dated profiles for static frontends and server-side AI backends, including streaming, execution limits, and secrets; package the supplied mock AI application with recorded prompt/model-configuration versions.

Module 2: Controlled AI release and exposure

Serverless deployment: Compare model-call duration, streaming, background work, and ephemeral state against supplied runtime limits; separate application release from durable agent tasks and external effects.
Tunnelling: Specify access restrictions, expiry, and cleanup evidence for simulated exposure; verify no local exposure process or simulated permission remains.
Preview environments: Version application, prompt, and model configuration together; use synthetic data, checked approval, and access cleanup without assuming release recovery reverses completed tool effects.

Practical outcome: Run supplied contract/evaluation gates in a local promotion simulator; bind approval to the artifact and configuration and inspect recovery and preview-access cleanup.

Practical exercise

Submit a mock AI web build, hosting decision, versioned release ledger, and promotion/recovery check results.
Test valid release, failed evaluation fixture, model configuration changed after approval, repeated promotion, and injected release failure.
Reject failed-check or changed-release promotion; keep dummy provider secrets out of browser artifacts and verify one release effect for repeated promotion.
Restore the prior artifact and configuration and verify local/simulated exposure cleanup; list data changes and tool effects that release recovery cannot undo.

What you will produce

Promote and recover a versioned AI web artifact and configuration through checked approval, server-side secret handling, and exposure cleanup.

How the work is checked

Failed contract/evaluation checks and artifact/model-configuration changes prevent promotion; dummy provider secrets are absent from browser builds.
Repeated promotion has one release effect and failure recovery restores the recorded artifact.
The ledger records artifact identity, approval, checks, and verified exposure cleanup.

Four-hour vendor-neutral AI web release simulation with supplied quality gates and hosting profiles. Do not equate fixture evaluation with production model quality. No live provider deployment, public tunnel, or advanced streaming implementation is promised; restoring an artifact/configuration cannot reverse external data changes or completed tool effects.

When another approach is better

Simple static hosting is preferable when no server runtime or complex release path is needed.
Do not open a tunnel when access cannot be limited and cleanup verified.
A local release simulation cannot establish a vendor's runtime, availability, or rollback guarantees.

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