Build the foundations
Build an API application and understand model inputs, outputs, tokens and structured responses.
Loading your career tools…
Move from a prompt demo to a tested AI application.
Build an API application and understand model inputs, outputs, tokens and structured responses.
Implement retrieval with document permissions, chunking and citations; add tool use only where needed.
Create evaluation cases for correctness, refusal, injection and leakage. Track latency and cost.
Deploy a small assistant with a human escalation route and documented failure cases.
Your checklist stays in this page session only. It is a self-check, not a skills assessment.
Choose the cloud and application stack your target employer uses. Claude certifications currently require partner access; public Claude learning is a separate option. Check every exam’s current availability.
A direct fit for developers creating and deploying AI apps and agents in Microsoft Foundry.
Before booking: Be comfortable developing Python applications and working with Azure and generative AI.
A platform-specific route for building generative AI solutions around Databricks data and AI services.
Before booking: Practise Python, retrieval pipelines, model serving and application evaluation.
A strong specialisation for developers building production generative AI applications on AWS, including secure retrieval, agent workflows and operational quality.
Before booking: AWS targets 2+ years building production applications and one year implementing generative AI solutions.
Demonstrate Claude application-building skills.
Before booking: For engineers with coding and API experience.
Review the official source for fees, assessment format, prerequisites and award/renewal requirements. Guidance reviewed 13 September 2026.
Build a policy assistant over public documents with citations, access boundaries and a versioned evaluation set.
Use these original practice scenarios to structure your answers. They are not leaked exam questions or an employer’s interview script.
A strong answer covers: Compare retrieval and answer evaluations on representative cases, including regressions and cost.
A strong answer covers: Treat retrieved text as untrusted data, restrict tools and test injection scenarios.