AWS AI and Machine Learning Certification Path

AWS now has a distinct AI certification landscape that spans foundational AI knowledge, machine-learning engineering, generative-AI development, and business strategy. Current AWS certifications no longer fit the old idea that machine learning on AWS is represented by one specialist exam. Candidates need to decide whether they are learning to understand AI, engineer AI systems, build generative-AI applications, or lead AI adoption.

The path is also changing quickly. AWS opened the MLA-C02 beta in September 2026, and the last day to take MLA-C01 in English was September 28. The older MLA-C01 remains available in several non-English languages during the transition, while general availability of MLA-C02 is planned for January 2027. That means English-language candidates preparing in October 2026 should not treat MLA-C01 as the default current exam.

AWS also introduced the AI Business Strategist beta in September 2026, creating a non-implementation route for professionals responsible for AI investment, governance, adoption, and business outcomes. That new credential sits alongside rather than above the technical certifications.

AIF-C01 is the technical-literacy foundation

AIF-C01 supports AWS Certified AI Practitioner and is the clearest foundation for candidates who need structured knowledge of AI, machine learning, generative AI, responsible AI, and the AWS services used to support those capabilities. It is designed for broader technical literacy rather than deep implementation.

AIF-C01 AI fundamentals are useful for people who work with AI projects but do not yet own production model pipelines: cloud professionals, analysts, product leaders, developers beginning to add AI features, and technical sellers. The goal is to understand what different AI approaches can do, what data and model concepts matter, and where risks such as bias, hallucination, privacy, security, or inappropriate use appear.

A foundational credential has the most value when it helps candidates ask better questions. Can the problem be solved by rules instead of AI? Does the use case need predictive ML or generative AI? Is retrieval enough, or is model adaptation required? What data is available, and can it legally and safely be used? Those decisions matter long before an engineer chooses an AWS service.

AWS AI Business Strategist creates a separate business route

AWS Certified AI Business Strategist (AIB-C01) is a new beta credential for professionals who decide which AI initiatives should be funded, how value should be measured, what governance should apply, and how an organization moves from pilots to repeatable adoption. It is not positioned as a configuration exam and does not expect candidates to administer AWS infrastructure.

That makes it different from AI Practitioner even though both can serve non-specialist audiences. AI Practitioner validates technical understanding of AI and AWS AI concepts. AI Business Strategist validates business judgment: identifying appropriate opportunities, building a credible case, assessing readiness, managing change, setting responsible-AI governance, and deciding when a project should scale.

Because AIB-C01 is a new beta credential, candidates should rely on AWS’s current exam guide and beta information rather than assume older AI certification maps already include it. The business route is now part of the live AWS certification landscape even though much existing third-party study material still focuses on the earlier technical credentials.

MLA-C01 is in transition; MLA-C02 is the current English beta

MLA-C01 and the AWS Certified Machine Learning Engineer – Associate credential remain useful for candidates studying the established machine-learning engineering scope. However, AWS ended English delivery of MLA-C01 on September 28, 2026 and began MLA-C02 beta delivery on September 29.

MLA-C02 reflects how the ML engineer role has expanded. Traditional data preparation, model training, deployment, monitoring, and operations remain important, but the updated exam adds stronger coverage of generative AI, foundation models, retrieval-augmented generation, agentic workflows, Amazon Bedrock, and responsible AI. That shift mirrors production reality: many ML engineers now support both classical models and generative-AI systems.

Historical coverage of MLA-C01 machine learning engineering can still teach durable concepts such as data quality, model evaluation, deployment, observability, and security. English-language candidates should nevertheless build their exam-specific preparation around MLA-C02. Because MLA-C02 is still in beta, candidates should verify current registration and study-guide details directly with AWS rather than treating MLA-C01 material as the default current target.

Machine-learning engineering is broader than model training

A strong ML engineer on AWS needs to understand the lifecycle around models. Amazon SageMaker is part of that conversation, but the work extends from data preparation and feature pipelines through experiments, model selection, deployment, monitoring, automation, security, and cost control.

The engineering challenge is repeatability. A model that performs well once in a notebook is not yet a production system. Teams need versioned data and artifacts, controlled environments, evaluation criteria, reproducible deployment, observability, rollback mechanisms, and a clear response when quality changes. Generative-AI systems add new variables such as prompts, retrieval sources, foundation-model versions, guardrails, tool behavior, and agent workflows.

Candidates preparing for the updated associate route should therefore practice tracing a system end to end. If an answer quality metric drops, is the cause the model, the retrieval corpus, a prompt change, a data pipeline, a tool, or a monitoring error? Production ML engineering is the discipline of making those causes visible and manageable.

AIP-C01 is the professional route for generative-AI builders

AIP-C01 supports AWS Certified Generative AI Developer – Professional. It targets developers who integrate foundation models into applications and business workflows and validates production-oriented generative-AI skills rather than classical model development.

The exam scope includes foundation-model integration, retrieval-augmented generation, vector stores and embeddings, agentic systems, tool integration, safety, security, governance, testing, validation, performance, observability, and operational efficiency. This makes the credential much closer to application architecture and AI platform engineering than to a prompt-writing certificate.

Candidates should understand Amazon Bedrock as a platform for integrating foundation models and generative-AI capabilities, but they should avoid reducing preparation to a service-feature checklist. A production GenAI developer needs to decide how information enters the system, how models are selected, how tools are exposed, how outputs are evaluated, how sensitive data is protected, and how the application behaves when a model gives an uncertain or unsafe result.

Safety and governance belong inside the technical design

Generative AI raises operational risks that cannot be bolted on at the end. The professional developer route explicitly includes safety, security, and governance because an application may process sensitive input, retrieve internal data, produce regulated content, or call tools that have real-world effects. Amazon Bedrock Guardrails is one example of the controls teams can use, but guardrails are only one layer.

A mature design also considers IAM, data isolation, encryption, logging, prompt injection, tool authorization, content filtering, model evaluation, human review, and incident response. If an agent can modify a system, the application needs a stronger control model than a read-only summarization tool. If the output influences a financial, legal, or clinical process, the evaluation and escalation requirements are also different.

This is where certification study should connect directly to architecture practice. Candidates can take one use case and ask what could go wrong at every boundary: user input, retrieved data, model behavior, external tools, stored state, output handling, and downstream automation. That exercise builds the judgment that professional-level exams are trying to measure.

The certifications represent different responsibilities, not a single ladder

AIF-C01 is a good foundation for people who need AI literacy. AIB-C01 is aimed at leaders and business professionals deciding how AI creates value and how adoption should be governed. MLA-C02 is the current English transition route for engineers operationalizing machine learning and generative AI. AIP-C01 is for experienced developers building production generative-AI applications.

Those routes can overlap without needing to be sequential. A machine-learning engineer may already have enough foundational knowledge to skip AIF-C01. A business leader may take AIB-C01 without ever needing MLA or AIP. A senior developer may move directly toward the professional generative-AI credential if the experience requirements and day-to-day work justify it.

The most useful progression follows increasing accountability. Someone may begin by understanding AI, then become responsible for deploying models, then own a generative-AI application architecture. Another person may move from product management into AI strategy without becoming an implementation engineer. AWS’s expanded portfolio now supports both patterns.

Use the current exam version and preserve the durable concepts

The current AWS AI path includes AIF-C01, the machine-learning associate transition from MLA-C01 to MLA-C02, AIP-C01, and the new AI Business Strategist AIB-C01 beta.

Candidates should separate legacy study material from current certification status. MLA-C01 content still has value for durable ML-engineering concepts and for the remaining non-English transition period, but an English candidate in October 2026 needs MLA-C02-specific preparation. Likewise, AIB-C01 now belongs in a complete AWS AI path even though it is still in beta.

Hands-on practice should mirror the route. Foundation candidates can compare use cases and responsible-AI choices; ML engineers should build repeatable pipelines and investigate quality or deployment failures; generative-AI developers should implement retrieval, agents, evaluation, safety controls, and observable integrations. Business strategists should practice prioritizing use cases, defining measurable outcomes, and identifying the governance needed before a pilot becomes an enterprise capability. The certifications differ because the work differs.

The durable layer is the work itself: AI literacy, strategy, data quality, model and foundation-model integration, evaluation, responsible AI, security, deployment, monitoring, agent design, and cost management. Build those skills deeply, then use the current AWS exam guide to map them to the version that is actually available when you schedule.

• Category: All Certifications Certifications