March 31, 2026 was the final testing day for AWS Certified Machine Learning - Specialty (MLS-C01). You can no longer take or retake it, but an earned certification remains active for three years. This guide explains the historical scope, provides a worked model-evaluation exercise, and helps you check the separate current Associate path.
What retirement means for your credential
The last day to take the MLS-C01 exam was March 31, 2026. Existing certifications stay active for three years from the date they were earned.
AWS says existing holders need take no action because of retirement. You can continue displaying and sharing your Credly digital badge during its remaining three-year validity period.
Recertifying by retaking MLS-C01 was possible only on or before March 31, 2026. Check your certification record for its expiry date, and consider current learning or certification options when planning ahead.
Official sources: AWS Certified Machine Learning - Specialty — Certification page, retirement notice and exam overview
Use the historical syllabus to organize your notes
The old guide covered data engineering, exploratory data analysis, modeling, and ML implementation and operations.
The four weights below describe the retired Specialty exam. Its task and service lists were not exhaustive, and they are not the blueprint for the separate Machine Learning Engineer - Associate.
For each older lesson, identify the concept being taught, check current service behavior, and map it to your new goal. Keep useful explanations where they remain accurate; update the gaps using the guide for the exam you actually intend to take.
| Domain | Share of scored content | Task focus |
|---|---|---|
| Data Engineering | 20% | Data repositories for ML, ingestion solutions, transformation solutions |
| Exploratory Data Analysis | 24% | Sanitizing and preparing data, feature engineering, analysis and visualization |
| Modeling | 36% | Framing business problems as ML problems, model selection, training, hyperparameter optimization, evaluation |
| Machine Learning Implementation and Operations | 20% | Reliable ML solutions, service selection, security practices, deployment and operations |
Official sources: AWS Certified Machine Learning - Specialty Exam Guide (MLS-C01) — Exam guide PDF, content outline and scoring sections
Practice: compare a classifier with a simple baseline
A confusion matrix lets you calculate what the model gets right, which positive cases it misses, and how reliable its positive predictions are.
True positives are actual positives predicted as positive; false positives are actual negatives predicted as positive. False negatives are missed positives, and true negatives are actual negatives predicted as negative.
Accuracy counts all correct predictions. Precision measures how many predicted positives are actually positive; recall measures how many actual positives were detected. Choose measures that reflect the application’s goals and the costs of different errors.
The cited archived Amazon Machine Learning documentation supplies these stable metric definitions. This paper exercise does not require the legacy service or its console.
Practice exercise
An original study example evaluates a binary classifier on 200 validation cases at a fixed threshold. It produces 30 true positives, 10 false positives, 20 false negatives, and 140 true negatives. Calculate accuracy, precision, and recall. On the same cases, what accuracy would a baseline that predicts every case as negative achieve? Explain what the model’s accuracy does not tell you about missed positives.
Show answer
Model accuracy is (30 + 140) / 200 = 85%. Precision is 30 / (30 + 10) = 75%. Recall is 30 / (30 + 20) = 60%. The all-negative baseline would correctly classify the 150 actual negatives, giving 150 / 200 = 75% accuracy.
There are 50 actual positives and 150 actual negatives. The model detects 30 positives and misses 20, so it misses 40% of the positive cases despite 85% overall accuracy. Its precision shows that 30 of its 40 positive predictions are correct. The all-negative baseline gets every actual negative right and every actual positive wrong; its accuracy is therefore 75%, not the model’s 85%. Comparing the measures helps expose which errors overall accuracy can hide. These counts do not establish the best operating threshold for a different application.
Official sources: AWS Certified Machine Learning - Specialty Exam Guide (MLS-C01) — Exam guide PDF, content outline and scoring sections; Binary Model Insights - Amazon Machine Learning — Developer guide, evaluation metrics and score cut-off sections
Check the separate Machine Learning Engineer - Associate
AWS points candidates toward Machine Learning Engineer - Associate for implementing and operating ML workloads in production. It is a separate credential with its own exam guide.
As of September 14, 2026, MLA-C01 in English is still available through September 28. Registration for the updated MLA-C02 English beta opened September 1, with testing beginning September 29. Registration and delivery are different milestones.
MLA-C01 in Japanese, Korean, and Simplified Chinese remains available until MLA-C02 general availability; the official page lists that date as to be determined. Confirm your version, language, and appointment date before buying study material or booking.
The Associate focuses on production ML engineering and operations. Use its applicable current guide to assess your preparation; do not transfer the Specialty’s domain weights or assume your old badge converts automatically.
Official sources: AWS Certified Machine Learning - Specialty — Certification page, retirement notice and exam overview; AWS Certified Machine Learning Engineer - Associate — Certification page, exam overview and key dates
A checklist for choosing your next step
Choose a goal, confirm the current scope, and reuse older material selectively.
These are study suggestions. AWS does not require a specific training course, so choose resources that address the skills you need to learn or practice.
- If you hold the Specialty, record your existing badge’s expiry date.
- Decide whether your next goal is a current certification, practical ML engineering work, or deeper study of a particular topic.
- For certification, check the exact exam version, language, testing date, and current official guide.
- Map useful older lessons to the relevant current tasks and mark what needs updating.
- Use practice resources for the selected version and review the reasons behind wrong or guessed answers.
- Revisit the metrics exercise and explain the difference between overall accuracy, precision, recall, and the all-negative baseline.
Official sources: AWS Certified Machine Learning - Specialty — Certification page, retirement notice and exam overview; AWS Certified Machine Learning - Specialty Exam Guide (MLS-C01) — Exam guide PDF, content outline and scoring sections; AWS Certified Machine Learning Engineer - Associate — Certification page, exam overview and key dates; Binary Model Insights - Amazon Machine Learning — Developer guide, evaluation metrics and score cut-off sections; AWS Certification FAQs — Scheduling, scoring, and retake policy sections
Who the Specialty was designed for
AWS described its target candidate as someone with at least two years of experience developing, architecting, and running ML or deep learning workloads on AWS.
Recommended background included relevant frameworks and basic hyperparameter optimization. These were preparation recommendations. AWS did not require a particular certification first.
The historical role excluded specified advanced tasks such as complex algorithm development and mathematical proofs. Use a current exam’s own role description when deciding what depth to study next.
Official sources: AWS Certified Machine Learning - Specialty — Certification page, retirement notice and exam overview; AWS Certified Machine Learning - Specialty Exam Guide (MLS-C01) — Exam guide PDF, content outline and scoring sections
The historical MLS-C01 exam format
MLS-C01 used 65 questions in 180 minutes, with 50 scored questions and 15 unidentified unscored questions.
It used multiple-choice and multiple-response questions. The minimum passing score was 750 on a scale of 100–1,000, with an overall passing requirement rather than a separate pass in each domain. A scaled score is not a raw percentage correct.
The listed fee was USD 300 before applicable taxes and currency adjustments. These are historical details of the retired exam; they do not describe the format or fee of another AWS certification.
Official sources: AWS Certified Machine Learning - Specialty — Certification page, retirement notice and exam overview; AWS Certified Machine Learning - Specialty Exam Guide (MLS-C01) — Exam guide PDF, content outline and scoring sections
Official sources
Fact-checked against official AWS pages: