Red Hat OpenShift AI is Red Hat's platform for training, serving and monitoring models on OpenShift, the company's Kubernetes distribution. AI267, Developing and Deploying AI/ML Applications on Red Hat OpenShift AI, is the official course for it. The current version is based on OpenShift AI 3.3 and OpenShift 4.20. It is aimed at ML engineers handling MLOps and LLMOps work and at data scientists who train and deploy models. The outline is solid.
You start with workbenches and data science projects, move through model serving fundamentals and predictive model deployment, then monitoring, pipelines and advanced pipeline work, before finishing with GenAI optimisation and building GenAI applications. The exam objectives fill in the detail: KServe for serving, the model registry, TrustyAI and Grafana for monitoring, Elyra and Kubeflow for pipelines, LLM Compressor for shrinking models and building RAG apps. What I like is that this is the boring production work that actually decides whether an AI project survives. Most ML courses end when the notebook gets a good score. This one is mostly about what happens after that.
And because the platform is assembled from open source projects, a good chunk of what you learn applies on plain Kubernetes too. The EX267 exam is now called Red Hat Certified Developer in AI. Like other Red Hat exams it is performance-based: you get a live environment and have to make things work. Resellers list it at four hours and around $500. That is the right way to test this kind of skill, and Red Hat certifications have long been respected for it. The downsides are about who it is for.
The prerequisites are real. Red Hat recommends DO288 or equivalent OpenShift experience, plus Python and Git. If you have never touched OpenShift, you will spend the course fighting the platform. Pricing for the course itself is not published on the page and depends on region and format, which makes it hard to compare.
The Learning Subscription version, AI267LS, is 50% off until 31 December 2026 and bundles an exam attempt, which is the cheapest way in right now. There is also some naming confusion. The exam page address still says Specialist in OpenShift AI, and many training resellers still use that title. It is the same exam. Overall, this is a good, practical course for a narrow audience. If your organisation is on OpenShift and wants to run its own models, it is probably the most direct route.
Everyone else should learn the open source parts on their own.