Directory

An entry-level exam on AI, machine learning and generative AI concepts and the AWS services around them. It tests vocabulary and judgement, not the ability to build.

The original version of the ML Engineer Associate exam: building, deploying and operating classical ML workloads on SageMaker, with very little generative AI.

The rewritten ML Engineer Associate exam. It keeps the MLOps core and adds foundation models, RAG, vector databases, LLM evaluation and agent operations.

The AWS credential aimed squarely at developers building LLM applications in production: RAG, vector stores, prompt management, agents, guardrails and evaluation on Amazon Bedrock.

Formerly the deepest AWS machine learning exam, with over a third of the blueprint on modelling. It contains no generative AI content.

A business-role credential covering AI strategy, return on investment, governance and change management. Coding and implementation are out of scope.

A free, timed hands-on lab in a live AWS environment where the candidate fixes and extends agents. There are no multiple-choice questions.

Microsoft's entry-level AI exam, rebuilt around generative AI, agents and Microsoft Foundry. It is more technical than the AI-900 exam it replaced.

The earlier, purely conceptual version of Azure AI Fundamentals, covering machine learning, vision, language and generative AI at a descriptive level.

For years the main Azure credential for developers integrating AI services: vision, language, search and Azure OpenAI.

Azure's data science credential, covering training and running ML models with Azure Machine Learning.

Microsoft's current credential for developers building generative AI applications and agents in code on Microsoft Foundry, including RAG, evaluation and multi-agent orchestration.

An operations exam split between classical MLOps and GenAIOps: deploying foundation models, prompt versioning, evaluation metrics, tracing and RAG tuning.

A cloud back-end developer exam covering containers, databases, messaging and secrets. Its AI content is vector storage and similarity search; it has no prompting, model or agent objectives.

An agent-building credential for the low-code Copilot Studio and Power Platform stack. AI-103 is the code-first counterpart.

Microsoft's highest-level AI credential, aimed at solution architects working across Dynamics 365, Power Platform and Copilot Studio. It validates architecture judgement rather than coding.

An IT administration exam on licensing, data protection and monitoring for Microsoft 365 Copilot.

An end-user literacy exam for people who use Microsoft 365 Copilot in their daily work.

A credential for business decision-makers leading AI adoption, covering generative AI concepts at a vocabulary level.

Certifies proficient and responsible use of GitHub Copilot, including agent mode, MCP and organisation policies. It is about using an AI coding assistant, not building AI systems.

A short lab assessment in a real environment on building agents in Microsoft Foundry. Good evidence of hands-on familiarity with a narrow task.

Microsoft's expert-level credential for engineers who design, build and run production multi-agent systems on Microsoft Foundry. The blueprint also names MCP, A2A, LangGraph and LangChain.

An IT administration exam on configuring, securing and governing Microsoft 365 tenants and their AI services, including Copilot and agents. It does not test building AI applications.

A non-technical exam on generative AI concepts and Google Cloud's AI product range, written for business roles.

A senior ML engineering and MLOps exam on Google Cloud, with generative AI additions such as model selection, tuning and LLM evaluation. It does not test agent building.

A new professional exam on designing and running agentic systems on Google Cloud. It pairs a proctored exam with hands-on labs, a first for Google Cloud's professional tier.

A data platform exam. AI and ML make up under a tenth of the blueprint, and that part is about preparing data rather than building models.

An associate-level data exam with a small amount of beginner ML. Google Cloud has no associate-level certification dedicated to AI or ML.

Short lab-based badges for a single Google Cloud product skill, ending in a challenge lab. They are unproctored and are not certifications.

Roughly eight hours of beginner courses on using AI tools at work. Despite the name it is unrelated to Google Cloud's proctored Professional certifications.

A short AI literacy course on Coursera for a general audience.

A free education-sector exam on using the Gemini app, with versions for teachers and students. It is not a developer credential.

A hands-on coding exam in classical deep learning with TensorFlow. It has no generative AI content.

Anthropic's exam for engineers building on the Claude API, including tool use and agents.

The first Claude certification, for solution architects designing agent systems and production applications. It has the heaviest agent focus of the four Claude exams.

The top Claude tier, covering integration architecture, governance and evaluation at enterprise scale.

An end-user credential for consultants and project staff who use Claude in everyday work.

Free online courses with a quiz. They are separate from the proctored Claude Certified exams.

A badge for completing an OpenAI Academy course and a short unproctored quiz. It is not the formal OpenAI Certification.

OpenAI's planned proctored certification. It has no public exam guide, price or registration.

A live-proctored exam on building, testing, deploying and monitoring agents with LangChain, LangGraph and LangSmith. Documentation is allowed during the exam.

A free, self-paced course on building agents. The learner types their own name on the certificate.

Covers designing and deploying LLM applications on Databricks: RAG, chains, agents, serving and evaluation.

A new exam on what agents are given at inference time: prompts, retrieval, memory, MCP tools and context window management on Databricks.

Classical machine learning on Databricks: feature engineering, training, tuning, evaluation and deployment. It has no generative AI content.

Enterprise-scale ML and MLOps on Databricks, including distributed training, advanced MLflow and deployment testing. It has no generative AI content.

A one-hour entry-level exam that is mostly general machine learning and LLM theory rather than NVIDIA product knowledge.

A sibling of the LLMs Associate exam covering text, image and audio models.

One of the few professional exams dedicated to agents. Only a small share is NVIDIA-specific, which makes it unusually portable for a vendor certification.

Leans towards model-level engineering: optimisation, fine-tuning and GPU acceleration, more than application building.

Oracle's generative AI exam for developers on Oracle Cloud Infrastructure. Oracle has often offered it free during promotions.

A free, conceptual, entry-level exam. It is a foundations exam despite the word Associate in the title.

A free introductory exam on agents. Much of the syllabus covers non-Oracle technology such as LangChain and MCP.

Certifies low-code configuration of prompt templates and agents on Salesforce. It is not a general agent engineering credential.

Salesforce's former entry-level AI literacy credential.

Covers selecting, prompting and customising LLMs in watsonx.ai, with an unusually large share on fine-tuning for an associate exam.

IBM's agent credential: designing and managing agent orchestration on watsonx Orchestrate.

Generative AI on Snowflake, including Cortex AI functions, document processing and a large governance component.

A vendor-neutral security exam on protecting AI systems, including performance-based questions. It is written for experienced security professionals.

An expert-level, vendor-neutral data science exam covering classical ML, statistics and operations. It has no generative AI focus.

A two to three hour beginner course with a competency certificate. It is listed alongside CompTIA's certifications but is not one.

A Linux Foundation multiple-choice exam on the Model Context Protocol: building, securing and governing tool connections for agents.

A Linux Foundation multiple-choice exam on PyTorch fundamentals for early-stage practitioners. Unlike the foundation's Kubernetes exams, it is not hands-on.

The leading AI governance credential, covering law, standards and frameworks such as the EU AI Act, the NIST AI RMF and ISO 42001.

An advanced AI audit credential that can only be held alongside an existing senior audit certification.

An advanced AI security management credential for holders of CISM or CISSP.

An advanced AI risk credential for experienced IT risk professionals.

A project management credential for running AI projects, built on the CPMAI methodology. A preparation course is mandatory.

A vendor-neutral exam on classical machine learning: selecting, training, evaluating and implementing models.

A specialist testing credential on using generative AI in software testing. It is about applying AI to testing, not building AI products.

A 60-minute conceptual exam on generative AI, prompting, AI-assisted development and agents, including MCP. Despite the Cisco name, the blueprint mentions no Cisco products.

A planned vendor-neutral AI security certification from the issuer of CISSP, covering threats to AI systems alongside governance and risk.

DeepLearning.AI's well-known course series on neural networks, delivered on Coursera. It is course completion, not an exam.

A Coursera course series on machine learning and deep learning that now includes LLM and RAG content.

Paid online cohort courses from Stanford on AI topics. It carries no academic credit and is separate from the for-credit graduate certificate.

A series of short executive education courses. The certificate is essentially attendance-based.

A project-based course on building agent systems. Projects are reviewed by a person against a rubric, with unlimited resubmission.