Registry // Verified 13 September 2026
Every program below was checked against its official page. Each entry says what the credential actually proves, who it is for, and where the real learning path lives.
Read This First
Certifications prove familiarity with a defined body of knowledge. They do not prove you can build or judge a system in production. Programs also retire without warning, so confirm cost, format, and status on the official page before you pay.
Amazon Web Services
Every program listed on this page is an AI or machine learning credential. General cloud, data engineering, and platform certifications are deliberately excluded, even where the vendor groups them alongside AI.
Validates foundational knowledge of AI, machine learning, and generative AI concepts on AWS. A starting point for people who use AI services rather than build them. Ninety minutes, 65 questions, 100 USD, valid three years.
AWS names this as the next step after the AI Practitioner exam for people pursuing data, AI, and machine learning roles. Passing it also recertifies the AI Practitioner credential.
The most advanced AI credential in the AWS track, for developers building generative AI applications with AWS AI and machine learning services. AWS positions it alongside the Machine Learning Engineer exam as a progression from AI Practitioner.
Tests business judgment rather than implementation: evaluating AI investments, building the business case, designing governance, and scaling adoption. It does not assess AWS service knowledge and needs no coding experience. Currently a beta exam at 50 USD, 170 minutes, 85 questions.
Google Cloud
Covers building, evaluating, productionizing, and optimizing AI solutions on Google Cloud, including responsible AI practice. Two hours, 50 to 60 questions, 200 USD. Google recommends three or more years of industry experience.
A business-level credential for any role, technical or not. Covers generative AI fundamentals, Google Cloud's offerings, techniques for improving model output, and strategy for adoption. Ninety minutes, 50 to 60 questions, 99 USD, valid three years.
Microsoft Learn
The beginner entry point, covering AI concepts and capabilities alongside building AI solutions with Microsoft Foundry. Passing score is 700 and the exam carries no retirement date. Fundamentals certifications do not expire. Replaces AI-900.
The direct successor to AI-102, rebuilt around generative and agentic architectures. Covers planning and managing Azure AI solutions, generative AI and agentic solutions, computer vision, text analysis, and information extraction using Python and Microsoft Foundry. Two hours, proctored.
Covers MLOps and generative AI operations together, including model lifecycle, quality assurance, observability, and performance optimisation. Expects a data science background, Python, and entry-level DevOps with GitHub Actions and infrastructure as code. Replaces DP-100.
For developers building, integrating, and monitoring AI solutions on Azure using containerised compute, vector-enabled databases, event-driven pipelines, serverless functions, secret management, and distributed observability. Replaces AZ-204.
Extends the security engineer role to cover AI workloads: securing the platforms, data, identities, and infrastructure that AI systems depend on, alongside identity and access governance, storage, networking, compute, and security posture. Two hours, proctored. Replaces AZ-500.
NVIDIA
An entry-level credential covering the concepts behind developing and maintaining LLM applications: machine learning and neural network fundamentals, prompt engineering, alignment, experimentation, data preprocessing, Python libraries for LLMs, and deployment. One hour, 125 USD, valid two years.
Validates foundational knowledge of the compute, networking, and operations layer that AI workloads run on. One hour, 125 USD. Useful if you run the infrastructure rather than build the models.
The professional tier for people who deploy, configure, and validate advanced NVIDIA AI infrastructure, with separate exams for infrastructure, operations, networking, and rack and interconnect. Two hours each.
Vendor-Neutral Bodies
Credentials for people who must govern, audit, or defend AI systems rather than build them. These carry the most weight in regulated organisations and are not tied to any cloud platform.
The AIGP credential covers responsible AI principles, how current and emerging law applies to AI systems, the AI life cycle, and how risk is managed in context. The IAPP publishes a free body of knowledge and exam blueprint.
The AAIA certification covers AI governance and risk, AI operations, and AI auditing tools and techniques. It is designed exclusively for holders of CISA or another qualified ISACA designation, so check eligibility before buying anything. A free 12-question practice quiz is available.
ISACA also publishes AAISM for AI security management and AAIR for advanced AI risk. Both sit alongside its established CISA, CISM, and CRISC credentials.
CAIP validates engineering a dataset for use in a machine learning model, selecting algorithms, performing model training and handoff, and the ethics and oversight needed for responsible outcomes. It was the first AI certification accredited by ANAB to the ISO/IEC 17024 standard, which is a meaningful independence signal for a vendor-neutral credential.
Universities and Course Providers
These award course certificates rather than proctored certifications. They are the fastest honest way to build real understanding, and several are free.
A free course series from the University of Helsinki and MinnaLearn, taken by more than two million people across 170 countries. Introduction to AI needs no mathematics or programming. Building AI covers the underlying algorithms and expects basic Python.
A beginner-level specialization from DeepLearning.AI and Stanford Online, taught by Andrew Ng. The most widely recognised on-ramp to machine learning fundamentals.
Builds neural networks including convolutional, recurrent, and transformer architectures, and applies them to speech recognition and natural language processing.
A beginner course on what generative AI can and cannot do, alongside a large free library of short, focused courses on agents, retrieval, evaluation, and prompting.