How to choose an AI certification: role, platform, level
· 8 min read
Choose an AI certification by settling three things in order: the work you do or want to do, the platform your organisation already runs on, and how much hands-on experience you bring. Role narrows the field to a handful of exams, platform picks the provider, and experience picks the level. The table further down maps 23 current AI exams to those three decisions, with a study guide for each.
Start with the work, not the badge
Most AI certifications are written for a specific job. The provider's own description of the intended candidate is the most reliable signal of whether an exam fits you, so read it before anything else. Four broad groups cover almost every AI exam in our catalogue.
Practitioner and business roles
These exams test whether you understand what AI and generative AI can and cannot do, how to apply them to business problems and how to use them responsibly. You are not expected to build systems.
- AWS Certified AI Practitioner (AIF-C01). AWS describes it as foundational and aimed at anyone exploring AI, ML and generative AI on AWS, including people in business roles as well as cloud, development, data and IT.
- Google Generative AI Leader (GAIL). Google states it is for anyone in any job role, with or without hands-on technical experience.
- Claude Certified Associate: Foundations (CCAO-F). Aimed at professionals who use Claude for productivity work such as operations, marketing or project management, and explicitly not at developers building against the API.
One exam in this tier is more technical than its name suggests. Microsoft Azure AI Fundamentals (AI-901) is pitched at people at the start of a career in AI solution development, and Microsoft expects knowledge of Python syntax and familiarity with Azure resources. It suits a junior technologist better than a manager.
Developer roles
These exams assume you write code and integrate models, retrieval, tools and agents into applications. Examples include the AWS Certified Generative AI Developer – Professional (AIP-C01), Azure AI Apps and Agents Developer Associate (AI-103), Claude Certified Developer: Foundations (CCDV-F), Databricks Generative AI Engineer Associate, SnowPro Specialty: Gen AI, the LangChain Certified Agent Engineer, the Neo4j and Generative AI certification, GitHub Copilot (GH-300) and NVIDIA's two generative AI associate exams.
Architect roles
Architect exams test trade-off decisions: which pattern, which integration, which control, and why. The Claude Certified Architect exams (Foundations and Professional) and NVIDIA's Agentic AI Professional sit here. Cisco's AI Technical Practitioner (810-110) is broader and names solutions architects among its audience alongside technical IT professionals.
Machine learning engineering and MLOps
These exams cover the full model lifecycle: data preparation, training, deployment, pipelines and monitoring. The AWS Machine Learning Engineer – Associate (MLA-C02), Microsoft's Machine Learning Operations Engineer Associate (AI-300), Google's Professional Machine Learning Engineer (PMLE) and NVIDIA's Generative AI LLMs Professional belong in this group.
Then match the platform
Once you know your role group, choose the provider whose platform you use or expect to use. An exam that tests services you touch every week is easier to prepare for, and the result means more to an employer running that platform.
- Cloud platforms. If your organisation runs on AWS, Azure or Google Cloud, start with that provider's exams. Our [AWS vs Azure vs Google comparison](/blog/aws-vs-azure-vs-google-ai-certifications) sets their AI exams side by side.
- Data and application platforms. Salesforce, Snowflake, Databricks and Neo4j certify generative AI skills on their own products. These are the right choice when that product is where your AI work actually happens.
- Model and framework providers. Anthropic's Claude exams and the LangChain exam test how to build with a specific model family or framework, independent of which cloud hosts it.
- Hardware and model-stack depth. NVIDIA's exams concentrate on how models are built, trained, optimised and deployed, with less emphasis on any one cloud's managed services.
Then pick the level
Providers use different words for levels, so compare what they say about recommended experience rather than the label. A few examples from official pages:
- Google lists no recommended experience for Generative AI Leader.
- AWS recommends at least a year of experience with AWS machine learning services for the ML Engineer – Associate.
- For AIP-C01, AWS recommends two or more years building production-grade applications on AWS or with open-source technologies, plus a year of hands-on generative AI implementation.
- For PMLE, Google recommends three or more years of industry experience, including at least one year designing and managing solutions on Google Cloud.
If you meet the recommended experience for an exam, sitting it directly is reasonable. If you do not, a foundational exam from the same provider is a sensible first step because it teaches that provider's vocabulary and service names, which the higher-level exams then assume.
The 23 exams compared
Levels below are each provider's own label; "Not published" means the provider does not assign one. Each row links to our study guide, which carries the verified format, fee and domain weights.
| Exam | Provider | Level | Best fit | Study guide |
|---|---|---|---|---|
| AWS Certified AI Practitioner (AIF-C01) | AWS | Foundational | Business and technical newcomers on AWS | [AIF-C01 guide](/blog/aif-c01-study-guide) |
| AWS Certified Generative AI Developer – Professional (AIP-C01) | AWS | Professional | Experienced developers building generative AI on AWS | [AIP-C01 guide](/blog/aip-c01-study-guide) |
| AWS Certified Machine Learning Engineer – Associate (MLA-C02) | AWS | Associate | ML and MLOps engineers on AWS | [MLA-C02 guide](/blog/mla-c02-study-guide) |
| Azure AI Fundamentals (AI-901) | Microsoft | Fundamentals | Early-career technologists on Azure | [AI-901 guide](/blog/ai-901-study-guide) |
| Azure AI Apps and Agents Developer Associate (AI-103) | Microsoft | Associate | Python developers building AI apps and agents on Azure | [AI-103 guide](/blog/ai-103-study-guide) |
| Machine Learning Operations Engineer Associate (AI-300) | Microsoft | Associate | MLOps and GenAIOps engineers on Azure | [AI-300 guide](/blog/ai-300-study-guide) |
| Generative AI Leader (GAIL) | Foundational | Any role, no hands-on experience needed | [GAIL guide](/blog/gail-study-guide) | |
| Professional Machine Learning Engineer (PMLE) | Professional | Experienced ML engineers on Google Cloud | [PMLE guide](/blog/pmle-study-guide) | |
| Claude Certified Associate: Foundations (CCAO-F) | Anthropic | Foundations | Business users of Claude | [CCAO-F guide](/blog/ccao-f-study-guide) |
| Claude Certified Developer: Foundations (CCDV-F) | Anthropic | Foundations | Engineers shipping applications on Claude | [CCDV-F guide](/blog/ccdv-f-study-guide) |
| Claude Certified Architect: Foundations (CCAR-F) | Anthropic | Foundations | Solution architects building agentic systems | [CCAR-F guide](/blog/ccar-f-study-guide) |
| Claude Certified Architect: Professional (CCAR-P) | Anthropic | Professional | Senior architects leading enterprise deployments | [CCAR-P guide](/blog/ccar-p-study-guide) |
| Cisco AI Technical Practitioner (810-110) | Cisco | Not published | Technical IT staff and solutions architects | [810-110 guide](/blog/810-110-study-guide) |
| Databricks Generative AI Engineer Associate | Databricks | Associate | Engineers building RAG and LLM chains on Databricks | [Databricks guide](/blog/db-genai-assoc-study-guide) |
| SnowPro Specialty: Gen AI (GES-C02) | Snowflake | Specialty | Snowflake practitioners using Cortex | [GES-C02 guide](/blog/ges-c02-study-guide) |
| Salesforce Agentforce Specialist | Salesforce | Not published | Salesforce admins, developers and architects | [Agentforce guide](/blog/agentforce-spec-study-guide) |
| GitHub Copilot (GH-300) | GitHub | Intermediate | Developers using Copilot day to day | [GH-300 guide](/blog/gh-300-study-guide) |
| LangChain Certified Agent Engineer | LangChain | Not published | Engineers building and operating agents with LangChain | [LCAE guide](/blog/lcae-study-guide) |
| Neo4j and Generative AI Certification | Neo4j | Not published | Developers building GraphRAG applications | [Neo4j guide](/blog/neo4j-genai-study-guide) |
| NVIDIA Generative AI LLMs Associate (NCA-GENL) | NVIDIA | Associate | Entry-level LLM developers | [NCA-GENL guide](/blog/nca-genl-study-guide) |
| NVIDIA Generative AI Multimodal Associate (NCA-GENM) | NVIDIA | Associate | Developers working across text, image and audio | [NCA-GENM guide](/blog/nca-genm-study-guide) |
| NVIDIA Agentic AI Professional (NCP-AAI) | NVIDIA | Professional | Practitioners designing and deploying agents | [NCP-AAI guide](/blog/ncp-aai-study-guide) |
| NVIDIA Generative AI LLMs Professional (NCP-GENL) | NVIDIA | Professional | Practitioners training and fine-tuning LLMs | [NCP-GENL guide](/blog/ncp-genl-study-guide) |
You can browse all of them, with sample questions and syllabus coverage, on the [certifications page](/certifications).
Two worked choices
A product manager at a company that runs on AWS. Role group: practitioner. Platform: AWS. Level: no hands-on build experience. That points to AIF-C01. GAIL would test similar concepts but through Google Cloud's offerings, which the manager would not use at work.
A Python developer building agents on Azure, new to Azure AI. Role group: developer. Platform: Azure. Level: strong coding, little Azure AI exposure. AI-103 is the target, because Microsoft expects Python app development experience and familiarity with Azure services. Studying the AI-901 material first is optional; it is worth it only if the developer is also new to Azure itself.
Check timing before you commit
Exams change, and a few are changing right now.
- AWS is moving the ML Engineer – Associate from MLA-C01 to MLA-C02. According to the AWS certification page, the last English sitting of MLA-C01 is 28 September 2026 and MLA-C02 beta delivery begins on 29 September 2026.
- Google notes that the Professional Machine Learning Engineer exam was recently updated and directs candidates to the new exam guide.
- AWS lists the Italian and German versions of AIF-C01 as retiring after 15 October 2026.
Before you book, check the exam guide version on the provider's page and make sure your study material follows the same version.
Frequently asked questions
Which AI certification should a complete beginner take first?
If you do not write code, AIF-C01 or GAIL, chosen by whichever cloud your organisation uses. If you are starting a technical career and can read Python, AI-901 is a good fit on Azure.
Do I need a foundational certification before an associate or professional exam?
Not always. Google lists no prerequisites for GAIL or PMLE, for example, and many providers award credentials on exam performance alone. Check each exam page, and weigh the recommended experience more heavily than any formal requirement.
Are there vendor-neutral AI certifications in this list?
None of the 23 is fully vendor-neutral; each is issued by a provider and tests its own products to some degree. Cisco's 810-110 and the NVIDIA associate exams lean more towards general concepts than the cloud exams do.
How many AI certifications should I hold?
One that matches your current role is worth more than several that do not. Add a second when your work changes, for example moving from building applications to operating models in production.
Are CertifyAIx practice questions taken from real exams?
No. Every question is original and written against the provider's public exam guide. Our article on [how CertifyAIx practice questions are written](/blog/how-certifyaix-practice-questions-are-written) explains the method.
CertifyAIx is independent and not affiliated with any certification provider. All CertifyAIx practice questions are original.