Skip to content
Français
Account: not signed in, sync available My learning
Select theme
Français

CKIE: the Certified Kubernetes Inference Engineer certification

  • 15 min

By updated

Improve this page

Logo Kubernetes

The CKIE (Certified Kubernetes Inference Engineer) is a CNCF certification expected in 2027. The CNCF named it on November 10, 2026, at KubeCon + CloudNativeCon North America, when it launched the CKNE. Its name suggests that it covers serving AI models on Kubernetes, but the CNCF has not confirmed this. As of November 14, 2026, it has no curriculum, no price, no format and no exact date.

This page says what is official, what is not yet, and how to work on the useful basics from now on. It is for people who already administer or understand Kubernetes and want to serve AI models on it.

What you will learn

  • Say what the CNCF has announced about the CKIE, and when.
  • Spot what is not published yet as of November 14, 2026.
  • Place the CKIE among the other Kubernetes certifications.
  • Choose where to start preparing.

What is the CKIE?

The name of the CKIE, Certified Kubernetes Inference Engineer, means an engineer certified in inference on Kubernetes. Inference is the step where an already trained AI model answers requests, as opposed to training, which builds it.

The CNCF has not described the certification in more detail. The name suggests that it targets serving AI models on Kubernetes. This is a deduction from the name, not information given by the CNCF.

What do we know exactly, as of November 14, 2026?

The only official source is the CNCF announcement of November 10, 2026, published at the launch of the CKNE in Salt Lake City. It contains this sentence: "Looking ahead, CNCF also plans to introduce two additional certifications in 2027: Kubernetes AI Associate (KAIA) and Certified Kubernetes Inference Engineer (CKIE)."

It adds: "These upcoming credentials will further expand CNCF's training and certification portfolio as AI and cloud native technologies continue to converge." The table sums up what is known about the CKIE from this announcement.

InformationStatus as of November 14, 2026
NameCertified Kubernetes Inference Engineer (CKIE)
Announced byCNCF, on November 10, 2026, at KubeCon + CloudNativeCon North America
Timeframe2027, no exact date
Curriculum (domains and weights)Not published
PriceNot published
Exam formatNot published
PrerequisitesNot published

As long as the CNCF and the Linux Foundation have published nothing, any other figure or detail found elsewhere about the CKIE has no source.

Where does the CKIE fit among the Kubernetes certifications?

The CKIE would join the CNCF Kubernetes certifications, alongside the KAIA announced at the same time. Six certifications already exist, the CKNE being the most recent.

CertificationTopicFormat
KCNAKubernetes and cloud native fundamentalsMultiple choice
KCSACloud native security fundamentalsMultiple choice
CKACluster administrationHands-on exam
CKADApplication developmentHands-on exam
CKSCluster securityHands-on exam
CKNECluster networkingHands-on exam
KAIAAnnounced for 2027Not published
CKIEAnnounced for 2027Not published

The guide to choosing your Kubernetes certification compares the six existing certifications. The page on the CKNE gives an example of what a published curriculum brings: five weighted domains, a format and a price. For the CKIE, nothing of the sort exists yet.

What is the difference between the CKIE and the KAIA?

The CNCF has given no detail about the difference. It announces both together, for 2027, without describing their audience or their content. Only the names allow a hypothesis.

On the CKIE side, the name points to an engineer role on a specific topic, inference, as the CKNE does for networking. The word Associate in the KAIA recalls the entry-level exams KCNA and KCSA; the KAIA page details this comparison. These comparisons remain a hypothesis of this page's author: neither the level nor the format of these exams is published.

How do you prepare for the CKIE right now?

Without a curriculum, the most useful move is to consolidate two foundations: Kubernetes itself and how model inference works. Neither will be wasted, whatever the final curriculum.

The Kubernetes foundation is that of the CKA: administering a cluster, deploying applications, diagnosing a failure. An inference service is still a Kubernetes application, and you must know how to deploy and troubleshoot it before dealing with models.

The inference foundation is built with the site pages on the topic:

  1. Understanding LLM inference (in French): the prefill and decode phases, the KV cache, the difference between latency and throughput.
  2. Serving an LLM (in French): exposing a model behind an API.
  3. vLLM (in French) and SGLang (in French): two inference engines to know.
  4. The comparison of inference backends (in French): choosing between Ollama, llama.cpp, vLLM, SGLang, TGI and LMDeploy.
  5. AI with Python (in French): the entry point of this series of pages.

These pages are not the CKIE curriculum, which does not exist yet. They teach notions that an exam on serving AI models is likely to assume you already know.

Which Kubernetes topics could matter for the CKIE?

This list is an author's hypothesis, deduced from the name of the certification. It is not the CNCF curriculum, and it may differ.

Serving a model on Kubernetes raises placement and resource questions. A model often runs on a node equipped with a GPU, a compute card much faster than a processor for this type of workload. Two distinct steps then come into play.

The first is to request a GPU. The Pod states it in its resources, the same way it requests memory. For Kubernetes to know that the node has this GPU, a device plugin must run on it: a small program that declares the hardware to the cluster. The guide Talos Linux with CSI, CCM and GPU on OUTSCALE (in French) shows one, the NVIDIA one.

The second is to reserve those nodes for the workloads that need them, because a GPU is expensive. You put a taint on them, a mark that repels Pods. Only Pods that carry the matching toleration, that is, the permission to ignore this mark, can be placed there. The page on advanced scheduling covers this with taints, tolerations and affinity.

For serving the models themselves, the Kubeflow (in French) guide presents machine learning pipelines on Kubernetes. It briefly describes KServe, the project of the Kubeflow ecosystem that serves models.

Inference traffic could matter too. The CKNE, announced the same day and whose curriculum has been published since October 2026, already includes an objective named Optimizing LLM Traffic, in its Advanced Traffic Management domain. The CKIE is not the CKNE, but this example shows that the CNCF is already bringing Kubernetes and AI together in its curricula.

FAQ

When will the CKIE be available?

The CNCF announces the CKIE for 2027 and gives no more precise date as of November 14, 2026. The announcement of November 10, 2026 only says that the CNCF plans to introduce the CKIE and the KAIA in 2027.

How much will the CKIE cost?

The price is not published as of November 14, 2026. No source allows you to deduce it from the price of the other certifications.

What is the CKIE curriculum?

It does not exist yet: no domains, no weights and no list of objectives are published as of November 14, 2026. The topics mentioned in this page are the author's hypotheses, based on the name of the certification.

What is the difference between the CKIE and the KAIA?

The CNCF announces the CKIE and the KAIA together for 2027 without describing them. Their differences in audience, level and format are not published. You have to wait for the official curricula to compare them.

Where should you start to prepare for the CKIE?

Start with the Kubernetes basics of the CKA, then read the site pages on LLM inference (in French) and on serving an LLM (in French). These are useful foundations whatever the final curriculum.

Key takeaways

  1. The CKIE, Certified Kubernetes Inference Engineer, is announced by the CNCF for 2027.
  2. The announcement dates from November 10, 2026, at the launch of the CKNE, at KubeCon + CloudNativeCon North America.
  3. As of November 14, 2026, no curriculum, price, format or exact date is published.
  4. The "inference" topic is a deduction from the name, not information from the CNCF.
  5. The CNCF also announces the KAIA for 2027, with no further details.
  6. To prepare, consolidate the CKA basics and the notions of LLM inference.