Hello, you are using an old browser that's unsafe and no longer supported. Please consider updating your browser to a newer version, or downloading a modern browser.
Everything you need to know about the Certified Artificial Intelligence Practitioner (CAIP), the vendor-neutral machine learning certification from CertNexus, as of October 2026. This guide covers the AIP-210 exam, the four domains and their weights, the question formats and passing score, the $367.50 voucher, ANAB accreditation, the roles it fits, how it compares with the AWS AI Practitioner and CompTIA Data+, and how to renew it.
The Certified Artificial Intelligence Practitioner (CAIP) is the practitioner-level AI certification from CertNexus. It tests whether you can take a business problem, turn it into a machine learning task, prepare the data, train and tune a model, and put that model into production without leaning on any one cloud vendor’s tools.
CertNexus describes a certified practitioner as a data professional who can engineer the features of a dataset, select algorithms, train a model and hand it off, and who understands the ethics and oversight AI work needs. Exam AIP-210 launched in March 2023 and follows blueprint version 1.10. Ethics runs through the whole exam as the closing objective of all four domains, so a question about feature engineering can just as easily turn out to be a question about bias or privacy.
The credential carries no prerequisites and no application fee, but the blueprint recommends 1 to 3 years of professional exposure along with statistics, Python or R, and a querying language. It is issued and maintained by CertNexus, a Rochester, New York certification body that is part of Logical Operations, and CertNexus states that AIP-210 is accredited by ANAB under ISO/IEC 17024.
Business use cases, ranking ML approaches, stakeholder communication and ethical concerns. 26%.
Data quality, transformations, data formats and the risks inside feature engineering. 20%.
Model design, algorithm optimization, data splits and model evaluation. 24%.
Deployment, pipeline security and maintenance after production. The largest domain at 30%.
Four things that set CAIP apart from the cloud vendor exams and the AI literacy badges it usually gets compared with.
Many AI certifications either come from a cloud vendor or test AI awareness. CAIP does neither. CertNexus places it at the practitioner level of its AI line, a step past the AIBIZ and GenAIBIZ business badges, and the AIP-210 target candidate is someone who can design, implement and hand off an ML solution. The blueprint doesn’t name a single cloud service, so what you learn carries over to whichever platform your employer runs.
CertNexus states that AIP-210 is accredited by the ANSI National Accreditation Board (ANAB) under ISO/IEC 17024, the standard for personnel certification. That separates it from a course-completion certificate, because the exam is developed, scored and maintained under an audited process, with subject matter experts setting the cut score.
Operationalizing ML Models carries 30 percent of AIP-210, more than any other domain. Deploying a model, securing the pipeline and maintaining it after release outweigh training it. Our rundown of every major AI certification in 2026 shows where a build-focused credential sits against the governance and security ones.
CertNexus does not claim a DoD 8140 qualification for CAIP, so don’t plan on it to satisfy a cyber-coded position requirement on its own. The current matrix is published at the DoD Cyber Exchange. ANAB accreditation is one of the two tests DoD applies, but an alignment score against a specific work role is the other, and accreditation alone puts nothing on the matrix.
The recognition CAIP does have is specific. CertNexus says it is approved for funding through the military COOL programs, where each branch sets its own rules. The AIP-210 blueprint was written with subject matter experts from the National Security Agency, Booz Allen Hamilton, Delta Air Lines, the Centers for Disease Control and Prevention and the University of Waterloo, which is a fair signal of who the exam is built to satisfy.
Everything you need to know about the certification, the AIP-210 exam, and how to keep it current, as of October 2026.
CAIP sits in the middle of an ML career, not at the start. From here most people go one of three ways: further into building with AI agents, onto a cloud platform’s ML exam, or across into securing AI systems.
The next build credential from the same body. CertNexus’s Agentic AI Builder (AGB-110) moves from training models to designing, testing and deploying AI agents that reason, call tools and pull knowledge through retrieval-augmented generation. It is an online exam delivered through CertNexus’s CHOICE platform.
The platform step. CAIP never asks which managed service deploys a model, and an employer that runs on one cloud will want that too. On AWS the matching exam is the Machine Learning Engineer Associate (MLA-C01), which covers preparing data, training, deploying and monitoring models on AWS. Our AWS AI Practitioner guide lays out the rest of that track.
The security step across. CompTIA SecAI+ (CY0-001) launched February 17, 2026 and tests securing AI systems and using AI in security operations, in up to 60 questions over 60 minutes. For a CAIP holder whose models now go through a security review, it picks up where objective 4.2, Secure a pipeline, leaves off.
CAIP sits at the practitioner tier. It builds on data and programming groundwork rather than on another certification, and from there the path splits toward AI agents, AI security or AI governance.
The data and code groundwork
Building and deploying models
Two questions to answer before you commit: can you sit it, and should you pursue CAIP specifically. Here’s a straight answer to both.
No application fee and no required training.
CertNexus asks for no application fee, supporting documents or eligibility check before AIP-210. You buy a voucher, register through Pearson VUE, and pick a test center or an OnVUE online session. Training is optional as well. CertNexus says candidates can prepare through authorized training, self-study, prior education or work experience, and it publishes a readiness assessment if you want a baseline first. The blueprint still recommends 1 to 3 years of professional exposure, and that recommendation is worth taking seriously.
Simple item formats, sitting on top of a heavy recommended background in math and code.
The blueprint’s recommended background reads like a data science job posting: linear algebra, probability, multivariate calculus, distributions such as Poisson, normal and binomial, programming in Python or R, ensemble methods like random forest and XGBoost, and a querying language. Items can include code snippets, SQL among them, and data visualizations you have to read. CertNexus’s own course outline expects comfort with Python, NumPy and pandas before the first lesson. If you come from the business side of AI, budget real time for the math and the Python, because the problem-framing domain is only 26 percent of the exam.
CertNexus aligns CAIP with technical data roles, business and analytics roles, and a group of newer AI-enabled jobs. The six below are the ones where the AIP-210 domains match the daily work most closely.
Takes a trained model and makes it work in production. Domain 4, the largest on the exam at 30 percent, is this job on paper: deploying the model, securing the pipeline and keeping the model healthy after release.
Frames the problem, prepares the data and picks the algorithm. The first three AIP-210 domains follow that workflow in order, from ranking use cases by their chance of success to splitting data into training, validation and test sets.
Builds applications with models inside them. The exam’s coverage of deep learning model types, from CNNs to RNNs and LSTMs, together with deployment and pipeline security, gives a developer the model side of the conversation.
Already works with the data and wants to move into predictive work. Domain 2’s transformations, from normalization to log and logit transforms, and Domain 3’s evaluation work are the step from reporting on data to modeling it.
Builds the reports and data models a business runs on. CAIP adds the ML layer: judging which business problems a model can solve, and explaining results and risks to stakeholders, which objective 1.5 tests directly.
Turns research methods into working systems. AIP-210’s attention to model design, algorithm optimization, hyperparameter tuning and evaluation lines up with the applied side of research work.
CAIP gets weighed against two very different exams: the best-known vendor AI credential, and the vendor-neutral data analytics credential that often comes before it. Here’s how the three line up.
| CertNexus CAIP | AWS AI Practitioner | CompTIA Data+ | |
|---|---|---|---|
| Issuer | CertNexus | AWS | CompTIA |
| Level | Practitioner | Foundational | Early career |
| Focus | Building, training and deploying ML models | AI, ML, generative AI and the AWS AI services | Data analysis and reporting |
| Prerequisites | None (rec. 1 to 3 years) | None (up to 6 months AI/ML exposure) | None (rec. 18 to 24 months as an analyst) |
| Exam | 80 scored items, 120 min (AIP-210) | 65 items, 90 min (AIF-C01) | Up to 90 items, 90 min (DA0-002) |
| Domains | 4 | 5 | 5 |
| Passing Score | 60% (59% on some forms) | 700 of 1000 | 675 of 900 |
| Exam Cost | $367.50 | $100 | $264 |
| Validity | 3 years | 3 years | 3 years |
| Good Fit For | Data and development roles that build ML models | People who use AI on AWS in cloud, data, product and business roles | Early-career data analysts |
Prices are in USD as of October 2026 and can vary by region and tax. CAIP is the only one of the three that expects you to train and deploy models, and the only one scored as a straight percentage rather than on a scale. AWS AI Practitioner and Data+ facts reflect the Training Camp guides for those exams as of October 2026.
Certification Matchups: the published facts, strengths and weak spots of 33 certifications, side by side.
Our CertNexus CAIP boot camp covers all four AIP-210 domains in five days of live instruction, in person or online, with official CertNexus courseware, hands-on labs in a provided environment, exam prep questions and mock exams. Your exam voucher is included, Training Camp delivers CertNexus exams onsite, and exam assurance adds a free second voucher and a full course retake, with 180 days of one-on-one coaching after class.
Where CAIP sits among this year’s AI credentials, what ANAB accreditation does and doesn’t get you, and how to study from a blueprint like AIP-210.
A map of this year’s AI credentials, from the vendor exams that test building on one platform to the security and governance certifications, with a straight read on which job each one fits.
Why ISO/IEC 17024 accreditation is only one of the two tests that put a certification on the DoD 8140 matrix, and what the 70 percent alignment score adds.
The AIP-210 blueprint lists 20 objectives across four weighted domains. This is how to turn a document like that into a study plan instead of skimming it once.
How to read a practice exam the way an instructor does, starting with the domain breakdown instead of the total, and what a given score suggests about your next step.
How Army Credentialing Assistance requests work through ArmyIgnitED and Army COOL, and the fiscal year calendar and submission deadlines the process runs on.
A straight answer on the AI security credential that launched in February 2026, from someone who sat the beta, for practitioners deciding whether securing AI is their next step.
How the major certification bodies keep a credential current. Useful context for weighing CertNexus’s 90 credit, three year cycle against the renewals you already track.
The AIP-210 exam is organized into four domains with 20 objectives between them. Operationalizing ML Models is the largest at 30 percent, Understanding the Artificial Intelligence Problem follows at 26 percent, and every domain ends with an ethics objective. Click any domain for what it covers.
Where a machine learning project starts. The domain covers how AI and ML solve business problems in commercial, government, public interest and research settings, how to rank candidate ML use cases by their probability of success, and the business case for image recognition, natural language processing, speech recognition, predictive and recommendation systems, discovery and diagnostic systems, and robotics. It closes with two skills that get skipped in a lot of technical study plans: communicating with stakeholders and spotting ethical concerns before any data is touched.
The data work. You need to recognize how data quality and size change what an algorithm can do, explain the collection and transformation steps in an ML workflow (standardization, normalization, and log, square-root and logit transformations), work with textual, numerical, audio and video formats, and transform numerical and categorical data into something a model can use. The last objective puts business risk and ethics inside feature engineering itself, which is where many bias and privacy problems start.
Designing machine learning and deep learning models, telling the algorithm families apart, and designing for pattern recognition in predictive models. From there it moves to optimizing an algorithm through its structure, run time and hyperparameters, splitting data into training, validation and test subsets, and evaluating the result. The blueprint acronym list (CNN, RNN, LSTM, GAN, SVM, KNN, PCA, ROC, AUC and more) is a fair preview of the vocabulary you need here.
The largest domain, and the one that separates CAIP from the AI literacy exams. It covers deploying a model, securing the pipeline (maintenance included), and maintaining the model after it reaches production, plus the business risks and ethical concerns that come with a model running on live data. Candidates who study only algorithms and evaluation metrics leave almost a third of the exam to chance.
Domains and weights reflect version 1.10 of the CertNexus Certified Artificial Intelligence Practitioner (AIP-210) Exam Blueprint, modified March 5, 2025 and verified on October 21, 2026. Confirm the current blueprint on certnexus.com before scheduling.
The questions candidates ask most often when researching the CertNexus Certified Artificial Intelligence Practitioner.
The Certified Artificial Intelligence Practitioner (CAIP) is CertNexus’s vendor-neutral, practitioner-level AI certification. It validates that you can apply machine learning to business problems: framing the problem, engineering features, training and tuning models, and deploying and maintaining them in production, with ethics tested in every domain. It isn’t tied to any cloud platform, and CertNexus states that the exam is accredited by ANAB under ISO/IEC 17024.
The current exam is AIP-210. CertNexus launched it in March 2023 and lists its sunset date as TBD, so no replacement has been announced as of October 2026. This page follows version 1.10 of the AIP-210 exam blueprint, last modified March 5, 2025. Check the CAIP page on certnexus.com before you schedule, since CertNexus says its exam objectives can change without notice.
No. There are no prerequisites, no application fee and no eligibility check, and CertNexus doesn’t require training before the exam. The blueprint does recommend 1 to 3 years of exposure in a professional environment, along with applied math and statistics, programming for ML in Python or R, ensemble methods such as random forest and XGBoost, a querying language and familiarity with data visualization. If most of that is new to you, build it first.
The four domains are Understanding the Artificial Intelligence Problem (26 percent), Engineering Features for Machine Learning (20 percent), Training and Tuning ML Systems and Models (24 percent), and Operationalizing ML Models (30 percent). Deployment and production work is the largest single domain, so don’t spend all of your study time on algorithms. Each domain also ends with an ethics objective, which means ethics questions can turn up anywhere on the exam.
CertNexus lists 80 items, and the AIP-210 blueprint describes the exam as 80 scored items plus 10 trial items that don’t count toward your score. You get 120 minutes, but that includes 5 minutes for the Candidate Agreement and 5 minutes for the Pearson VUE tutorial, so plan on about 110 minutes for the questions themselves.
Multiple choice and multiple response. The blueprint says items are single-answer multiple choice by default, and that other item types may be used when the content calls for it, such as manipulating snippets of code (SQL included) or reading data visualizations. Expect to interpret code and charts, not only definitions.
You need 60 percent or 59 percent, depending on which exam form you get. CertNexus statistically equates its forms, so the one point difference adjusts for form difficulty rather than setting a different standard. If you pass, your score report shows only that you passed. If you don’t, it shows your performance in each domain so you can aim the retake.
The AIP-210 exam voucher is listed at $367.50 USD in the CertNexus store as of October 2026. Every CertNexus exam voucher includes one free retake, and vouchers expire 18 months after purchase. The exam is delivered at Pearson VUE test centers or online through Pearson OnVUE. Training Camp’s CertNexus CAIP boot camp includes the exam voucher, and its exam assurance adds a free second voucher and a full course retake.
CertNexus states that AIP-210 is accredited by the ANSI National Accreditation Board (ANAB) under ISO/IEC 17024. It does not claim a DoD 8140 qualification for CAIP, so don’t count on it to meet a cyber-coded position requirement by itself. CertNexus does list CAIP as approved for funding in the military COOL programs, where each branch sets its own eligibility rules. See the full DoD 8140 work role paths.
CAIP is valid for three years. You can renew by retaking and passing the current version of the exam before your certification expires, or by joining the CertNexus Continuing Education Program and earning 90 continuing education credits within the three-year cycle. The program fee is $150 per cycle, and it’s non-refundable. Related certifications, training courses, conferences and a year of related work experience can all count toward the 90 credits.
CAIP fits roles that build and run machine learning models: Machine Learning Engineer, Data Scientist, AI Developer, Data Analyst, Business Intelligence Developer and Applied Scientist, all of which CertNexus lists among the roles aligned with the certification. It shows an employer that you understand the full ML workflow without a vendor attached. Pair it with a portfolio of models you’ve built, because the exam is multiple choice and a portfolio shows the hands-on part.
They test different jobs. The AWS AI Practitioner (AIF-C01) is a foundational exam for people who use AI on AWS, and coding, feature engineering and hyperparameter tuning are out of its scope. CAIP is a practitioner exam for people who build models, and those topics sit at its center. If you work on AWS and don’t build models, take the AWS exam. If you build models, or work across several platforms, CAIP is the closer fit. Neither is a prerequisite for the other.
You won’t build a model from scratch during the exam. The item formats are multiple choice and multiple response, but the blueprint allows items that have you manipulate snippets of code, SQL included, and it recommends programming ability in Python or R. CertNexus’s own CAIP course outline expects comfort with Python, NumPy and pandas, so if you don’t have that background yet, learn it before you start exam prep.
If you’re weighing the certification, planning a move into machine learning work, or arranging AI training for a data team, tell us where you are and we’ll help you map out the right next step.