AI curriculum

Course outline for Cloudflare for AI Applications and Agents

An edge AI application needs separate roles for request handling, model inference, retrieval, and persistent agent state. Compare Workers, Workers AI, AI Gateway, and Vectorize with execution and coordination options. Simulate duplicate and concurrent requests while checking current limits rather than assuming every component has identical availability or isolation.

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

About Cloudflare for AI Applications and Agents

Cloudflare for AI Applications and Agents Course Objectives

Separate Workers request handling, Workers AI inference, AI Gateway control, and Vectorize retrieval.
Specify a persistent state owner and concurrent-update invariant for one AI request flow.
Run duplicate, stale-update, unavailable-model, and cache-data-scope fixtures.
Compare sandbox, container, and edge placement responsibilities.
Verify current product/model limits without assuming identical availability or isolation.

Pre-requisites

Basic JavaScript or TypeScript and HTTP knowledge.
Familiarity with application deployment and simple state updates.
Ability to inspect a small handler and its tests; a full cloud-platform course is not required.

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.

Recommended preparation

Readiness checks

Basic JavaScript/TypeScript and HTTP
Basic application-deployment familiarity

Can inspect or implement a simple HTTP handler; no full hosting-comparison or web-stack course is required for a local edge-state simulation. This is a provisional specialization.

Find your place in the learning map

Lab Setup

Node.js, code editor, local simulation scaffold, and dated official-reference profiles for Cloudflare AI, storage, and execution components.
Synthetic inference/retrieval, duplicate-update, cache-data-scope, and unavailable-model fixtures; no Cloudflare account, paid plan, or platform emulator is required.
A separate live product lab needs current official instructions, confirmed model/feature availability, and an approved cost/access plan.

Detailed Course Outline

Module 1: AI request handling and state ownership

Workers: Compare supplied request-handler profiles, execution limits, and remote dependencies; distinguish stateless request processing from ownership of persistent application state.
Durable Objects: Examine supplied state-coordination descriptions; specify an update invariant and compare duplicate requests with conflicting versions in the vendor-neutral simulation.
Dynamic Workers: Compare supplied dynamic-execution descriptions with fixed handlers; identify lifecycle, authority, tooling, and availability questions before choosing a live implementation.
Containers: Contrast packaged process runtimes with worker execution using supplied profiles; identify resource, lifecycle, and state responsibilities without assuming account availability.

Practical outcome: Annotate component roles for an AI request path and adapt a vendor-neutral state-owner simulation for duplicate requests and stale concurrent updates.

Module 2: AI services, placement, and execution boundaries

Sandboxes: Compare supplied isolation boundaries and permitted capabilities; distinguish execution containment from authorization and identify unverified platform properties.
Edge execution: Balance request proximity against remote-state coordination, network delay, dependency failure, and operational complexity; compare a simpler centralized placement.
Workers AI, AI Gateway, and Vectorize: Separate serverless model inference, provider proxy/control, and vector retrieval; compare model support, caching, rate limits, source/index ownership, and data paths using dated official profiles.

Practical outcome: Compare model inference, provider gateway, vector retrieval, and sandbox roles using dated profiles; inspect cache/data-scope and remote-coordination assumptions against a centralized alternative.

Practical exercise

Submit the annotated AI request simulation, invariant-test results, and service/placement decision record.
Run valid, repeated, conflicting-update, and unavailable-model cases; preserve protected state and use the documented timeout response.
Map source documents to R2, authoritative state to an appropriate store, and derived retrieval to Vectorize in supplied profiles; inspect a cache-sharing fixture for data-scope violations.
Distinguish Workers AI inference from AI Gateway proxy/control responsibilities and record current model, plan, runtime, and sandbox questions before proposing a live implementation.

What you will produce

Specify and test a simulated edge AI application while separating inference, gateway, retrieval, state ownership, and execution boundaries.

How the work is checked

Duplicate and stale-update tests preserve the declared state invariant.
The placement decision separates Workers AI, AI Gateway, and Vectorize roles and checks cache data scope against a centralized alternative.
Vendor assumptions and verification questions are explicit rather than presented as implemented features.

Provisional four-hour local simulation with dated Cloudflare profiles. Workers AI and AI Gateway have documented offerings, but individual models, limits, tooling, and execution features still need verification. No live inference, Vectorize deployment, sandbox assurance, or complete Cloudflare product lab is promised.

When another approach is better

A centralized application is simpler when edge placement adds coordination without a measured latency benefit.
Do not use a component whose verified availability, state semantics, or isolation properties do not fit the workload.

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