AWS Certified Machine Learning Engineer - Associate (MLA-C01)
Amazon Web Services
Complete guide to passing the AWS Certified Machine Learning Engineer - Associate (MLA-C01) exam on your first attempt.
$150
Not published
3 years
Global
Amazon Web Services
$112k-$141k
Are you ready for AWS Certified Machine Learning Engineer - Associate (MLA-C01)?
Loading quiz...
Complete Overview
The AWS Certified Machine Learning Engineer - Associate exam, exam code MLA-C01, tests whether you can build, deploy, and operate machine learning pipelines in the AWS Cloud, and one attempt costs 150 USD. Amazon Web Services aims it at backend software developers, DevOps engineers, data engineers, MLOps engineers, and data scientists who already have about one year of hands-on experience with Amazon SageMaker plus one year in a related engineering role. The exam runs 130 minutes and contains 65 questions. Fifty of those questions are scored and 15 are unscored trial items that AWS does not identify on screen. Results are reported as a scaled score from 100 to 1,000, and 720 passes. Scoring is compensatory, so a weak section can be offset by a strong one provided the overall scaled score clears 720. The exam guide numbers four content domains and fixes their weights: Domain 1, Data Preparation for Machine Learning, at 28 percent; Domain 2, ML Model Development, at 26 percent; Domain 3, Deployment and Orchestration of ML Workflows, at 22 percent; and Domain 4, ML Solution Monitoring, Maintenance, and Security, at 24 percent. Question formats include multiple choice with one correct answer out of four, multiple response with two or more correct answers out of five or more options, ordering items with three to five steps, and matching items with three to seven prompts. Ordering, matching, and multiple response items award no partial credit. Unanswered questions count as incorrect and there is no penalty for guessing. Pearson VUE delivers the exam at test centers and through online proctoring, with English proctoring available 24 hours a day, seven days a week. AWS offers MLA-C01 in English, Japanese, Korean, and Simplified Chinese, and candidates who register for a translated version can toggle to the English wording during the exam. The certification lasts three years. Holders renew by passing the latest version of the exam or by passing the AWS Certified Generative AI Developer - Professional exam, and the 50 percent discount voucher in the AWS Certification Account applies to either route. MLA-C01 took over the machine learning role from AWS Certified Machine Learning - Specialty, which AWS retired after March 31, 2026. AWS is now refreshing the associate exam itself: registration for MLA-C02 opens September 1, 2026, and the last day to take MLA-C01 in English is September 28, 2026. The Korean, Japanese, and Simplified Chinese versions of MLA-C01 stay available until MLA-C02 reaches general availability. Both versions target the same job role, so material written for MLA-C01 keeps most of its value, but read the published exam guide for the version code on your registration. AWS publishes no pass rate for any of its certification exams, so any percentage quoted elsewhere comes from a third party rather than from AWS. AWS does publish its standard-setting method: a panel of subject matter experts applies the modified Angoff technique to the first exam form, and every later form is statistically equated to that standard. The number of correct answers needed therefore shifts slightly between forms while the scaled cut score stays at 720. Preparation time tracks your SageMaker exposure. Engineers already running production pipelines often need six to eight weeks, and candidates arriving from general software or data work usually need ten to twelve.
Why Get AWS Certified Machine Learning Engineer - Associate (MLA-C01) Certified?
The fee is 150 USD, half of the 300 USD that AWS charges for its professional and specialty exams.
MLA-C01 is the only AWS associate certification dedicated to machine learning since AWS retired the ML Specialty exam after March 31, 2026.
AWS cites the World Economic Forum Future of Jobs Report 2023 on the exam page, which projects 40 percent growth in demand for AI and machine learning specialists.
The US Bureau of Labor Statistics projects 34 percent employment growth for data scientists from 2024 to 2034, against 3 percent for all occupations.
Passing MLA-C01 also renews AWS Certified AI Practitioner for three years, because AWS lists it as an accepted recertification option for AIF-C01.
Renewal can use the 50 percent discount voucher in your AWS Certification Account, which brings a three-year renewal down to 75 USD at current associate pricing.
AWS sets no formal prerequisites, so you can register without holding Cloud Practitioner, Solutions Architect Associate, or any other credential.
Exam Format & Structure
Duration
130 minutes
Questions
65
Passing Score
720 on a scaled range of 100 to 1,000
Question Types
- Multiple choice: one correct response and three distractors
- Multiple response: two or more correct responses out of five or more options
- Ordering: place three to five responses in the correct sequence
- Matching: pair responses with three to seven prompts
Delivery Method
Pearson VUE test center or Pearson VUE online proctored exam
- Fifty questions count toward your score and 15 are unscored trial items that are not identified during the exam.
- Ordering, matching, and multiple response questions require every element to be correct; AWS awards no partial credit.
- Unanswered questions are scored as incorrect and AWS applies no penalty for guessing.
- A Help button inside the exam lists the AWS service short names used in question text alongside their full names.
- Non-native English speakers taking the English version can request the ESL +30 accommodation for 30 extra minutes.
- Candidates sitting a translated version can switch individual questions to English with the on-screen language toggle.
How scoring works
Score scale
Scaled score from 100 to 1,000
Score needed to pass
720
Roughly what that means raw
AWS does not publish a raw cut score. Fifty of the 65 questions are scored, so roughly 40 correct scored answers is a sensible working target, but the exact raw requirement varies by form after equating.
When results arrive
Final results post to your AWS Certification Account within five business days of the exam, with an email notification. The Pearson VUE exit screen provides information about your results at the end of the appointment.
How the scale is built
AWS set the passing standard with the modified Angoff technique, in which a panel of content experts estimated the difficulty of each question against a minimally qualified candidate over multiple rating rounds. Every later exam form is statistically equated to that standard, so a harder form needs fewer raw correct answers. The equated raw score is then converted to the 100 to 1,000 scale. Scoring is compensatory, so no individual domain has its own pass mark.
- The score report includes a table classifying your performance at each section level, which AWS advises interpreting with caution because sections carry different question counts.
- Ordering, matching, and multiple response questions are scored all or nothing.
- Unanswered questions are scored as incorrect and there is no penalty for a wrong answer.
- Results can be delayed beyond five business days if they are held for security or technical review.
Pacing and time budget
65 questions in 130 minutes gives you 2 minutes per question per question.
| At this point | You should have answered |
|---|---|
| 26 min | 13 |
| 52 min | 26 |
| 78 min | 39 |
| 104 min | 52 |
| 120 min | 62 |
- The five checkpoints leave 10 minutes at the end for the flagged set, which is about five questions of rework.
- Long scenario questions describing an existing architecture take three to four minutes; short service-selection questions take under one, so the average holds if you keep moving.
- Ordering and matching items burn time through the drag interaction alone. Decide the sequence in your head before touching the interface.
- A timer warning appears when five minutes remain. Treat it as the deadline for converting blanks into guesses.
- Time spent typing in the Comment box is deducted from your 130 minutes, so leave feedback until after the review screen.
Where the marks are
| Topic | Weight | Why it scores |
|---|---|---|
| Data ingestion and storage selection across S3, EFS, FSx, RDS, and DynamoDB | 28% | Task 1.1 questions repeatedly ask which store fits a stated access pattern, throughput need, or cost ceiling. Knowing that Parquet suits columnar analytical reads and that Transfer Acceleration or Provisioned IOPS solve specific bottlenecks converts several questions in the largest domain. |
| Feature engineering with SageMaker Data Wrangler, Feature Store, and AWS Glue | 28% | Task 1.2 and 1.3 cover encoding, imputation, outlier treatment, class imbalance, and labeling. These appear as short scenarios where the right tool is decided by operational effort, and Data Wrangler versus Glue DataBrew versus Spark on EMR is a recurring comparison. |
| Choosing between built-in algorithms, managed AI services, and foundation models | 26% | Task 2.1 rewards recognizing that Rekognition, Comprehend, Textract, or Bedrock removes training work entirely. Candidates who default to a custom SageMaker model lose these questions, because AWS scores lowest cost and least operational overhead as the correct tradeoff. |
| Evaluation metrics, bias detection, and overfitting diagnosis | 26% | Task 2.3 asks which metric fits an imbalanced fraud or screening problem and how to read a confusion matrix, ROC curve, or SHAP output from SageMaker Clarify. The reasoning is arithmetic rather than AWS-specific, which makes these the most reliably winnable questions. |
| Monitoring, drift detection, and cost optimization | 24% | Tasks 4.1 and 4.2 cover Model Monitor baselines, CloudWatch alarms and dashboards, Inference Recommender, Compute Optimizer, tagging, Spot Instances, and Savings Plans. This domain outweighs deployment, and candidates who treat it as an afterthought give away 24 percent of the score. |
| IAM least privilege, VPC isolation, and KMS encryption for ML workloads | 24% | Task 4.3 questions usually turn on the scope of an IAM role or bucket policy rather than on encryption. Practicing a training-job role that reads exactly one S3 prefix, and placing an endpoint in a private subnet, makes these answers immediate. |
| SageMaker endpoint types, auto scaling, and compute selection | 22% | Task 3.1 and 3.2 test the real-time, serverless, asynchronous, and batch decision plus the auto scaling metric that fits, such as invocations per instance or model latency. It is a small set of rules that covers a large share of the deployment domain. |
| CI/CD orchestration with SageMaker Pipelines, Step Functions, and CodePipeline | 22% | Task 3.3 covers pipeline construction and rollout strategy. Knowing when Step Functions or Amazon MWAA beats SageMaker Pipelines, and matching blue/green, canary, and linear rollouts to risk tolerance, resolves the remaining deployment questions. |
The numbers
130 minutes, 65 questions
Exam length
Source: AWS Certified Machine Learning Engineer - Associate exam overview, aws.amazon.com/certification
50 scored questions plus 15 unscored trial questions
Scored versus unscored items
Source: AWS Certified Machine Learning Engineer - Associate (MLA-C01) Exam Guide, docs.aws.amazon.com
720 on a 100 to 1,000 scale, compensatory across domains
Minimum passing score
Source: AWS Certification After Testing policy page and the MLA-C01 exam guide
150 USD, or 128 EUR, 224 AUD, 20,000 JPY, 197,287 KRW, 1,057 CNY
Registration fee
Source: AWS Certification Before Testing exam pricing table
14 calendar days between attempts, no cap on total attempts, full fee each time
Retake wait and attempt limit
Source: AWS Certification FAQs and After Testing policy page
None. AWS publishes no pass rate for any certification exam
Published pass rate
Source: AWS Certification FAQ and MLA-C01 exam guide, neither of which reports a pass rate
If you fail
Wait before retaking
14 calendar days from a failed attempt
Attempt limit
No limit on the number of attempts
Retake fee
Full 150 USD registration fee for every attempt
The 14-day wait applies after each failed attempt, not only the first. Once you pass, AWS blocks a retake of the same exam for two years, unless AWS republishes it under a new exam guide and series code, which makes the new version available to you immediately. Beta exam takers may sit the beta version only once and must wait for general availability to retake it. Rescheduling is allowed twice per registration, and cancelling more than 24 hours before the appointment refunds the fee paid at purchase.
Exam Domains & Topics
The largest domain covers ingesting, transforming, and validating data before any model is trained. AWS tests three task statements here: ingest and store data, transform data and perform feature engineering, and ensure data integrity and prepare data for modeling. Expect questions that compare storage formats and services on cost, access pattern, and scalability rather than on raw capability.
Key Topics to Master:
- Data formats including Parquet, JSON, CSV, ORC, Avro, and RecordIO and when each fits an access pattern
- Ingestion from Amazon S3, Amazon EFS, Amazon FSx for NetApp ONTAP, Amazon RDS, and Amazon DynamoDB
- Streaming ingestion with Amazon Kinesis, Amazon Managed Service for Apache Flink, and Apache Kafka
- SageMaker Data Wrangler and SageMaker Feature Store for feature creation and reuse
- AWS Glue, AWS Glue DataBrew, and Spark on Amazon EMR for transformation at scale
- Encoding choices such as one-hot, binary, label encoding, and tokenization
- Outlier treatment, imputation, deduplication, scaling, binning, and log transformation
- Labeling with SageMaker Ground Truth and Amazon Mechanical Turk
- Class imbalance handling and dataset splitting for training, validation, and test
Tasks 2.1 to 2.3 move from choosing a modeling approach, to training and refining models, to analyzing model performance. AWS expects you to weigh a SageMaker built-in algorithm against a managed AI service such as Amazon Rekognition or Amazon Comprehend, and against a foundation model in Amazon Bedrock or SageMaker JumpStart, using cost, interpretability, and available data as the deciding factors.
Key Topics to Master:
- SageMaker built-in algorithms including XGBoost, Linear Learner, K-Means, and BlazingText
- Managed AI services: Amazon Rekognition, Amazon Comprehend, Amazon Transcribe, Amazon Translate, Amazon Bedrock
- SageMaker script mode with TensorFlow and PyTorch, and bringing external models into SageMaker
- Hyperparameter tuning with SageMaker automatic model tuning, random search, and Bayesian optimization
- Regularization with dropout, weight decay, L1 and L2, plus early stopping and distributed training
- Overfitting, underfitting, and catastrophic forgetting during fine-tuning
- Evaluation metrics: confusion matrix, precision, recall, F1, accuracy, RMSE, ROC, and AUC
- SageMaker Clarify for bias and explainability and SageMaker Debugger for convergence issues
- SageMaker Model Registry for versioning and reproducible experiments
- Ensembling, stacking, boosting, pruning, and compression to trade accuracy against size
The smallest domain runs Tasks 3.1 to 3.3: selecting deployment infrastructure, creating and scripting that infrastructure, and using automated orchestration tools to set up CI/CD pipelines. AWS wants you to match a workload to the right SageMaker endpoint type, then automate provisioning with CloudFormation or the AWS CDK and promote models through CodePipeline with a rollback strategy that fits the risk level.
Key Topics to Master:
- Real-time, serverless, asynchronous, and batch transform inference and the latency and cost tradeoffs
- Multi-model and multi-container endpoints for consolidating many small models
- Choosing compute for training and inference across CPU, GPU, and inference-optimized instances
- SageMaker endpoint auto scaling on model latency, CPU utilization, or invocations per instance
- Deployment targets beyond SageMaker: Amazon ECS, Amazon EKS, and AWS Lambda
- Infrastructure as code with AWS CloudFormation and the AWS CDK, including cross-stack references
- Containers with Amazon ECR and bring your own container for SageMaker
- Orchestration with SageMaker Pipelines, AWS Step Functions, and Amazon Managed Workflows for Apache Airflow
- CI/CD with AWS CodePipeline, AWS CodeBuild, and AWS CodeDeploy
- Blue/green, canary, linear, and shadow deployment strategies with rollback actions
Tasks 4.1 to 4.3 cover monitoring model inference, monitoring and optimizing infrastructure and costs, and securing AWS resources. Questions blend operations with governance: detecting data drift, spotting latency problems, tagging for cost allocation, and applying least privilege to training data, model artifacts, and endpoints. Security answers usually hinge on IAM policy scope and VPC isolation rather than on encryption alone.
Key Topics to Master:
- SageMaker Model Monitor for data quality, model quality, bias drift, and feature attribution drift
- Detecting distribution shift with SageMaker Clarify and running A/B tests in production
- Amazon CloudWatch metrics, logs, alarms, dashboards, and CloudWatch Logs Insights
- AWS X-Ray and CloudWatch Lambda Insights for latency troubleshooting
- AWS CloudTrail trails for auditing and for triggering retraining activity
- Rightsizing with SageMaker Inference Recommender and AWS Compute Optimizer
- Cost control with AWS Cost Explorer, AWS Budgets, Trusted Advisor, tagging, Spot Instances, and Savings Plans
- IAM roles and policies, S3 bucket policies, and SageMaker Role Manager for least privilege
- VPCs, subnets, security groups, and private endpoints for isolating ML workloads
- AWS KMS, AWS Secrets Manager, and Amazon Macie for protecting keys, credentials, and sensitive data
Recommended Study Plan
- 1Read the full MLA-C01 exam guide including the in-scope and out-of-scope service lists
- 2Enroll in the free Exam Prep Plan for MLA-C01 on AWS Skill Builder
- 3Create a dedicated AWS account with a budget alarm so lab spend stays visible
- 4Take the free 20-question Official Practice Question Set to get a baseline
- 5Write down every service in the in-scope list you cannot describe in one sentence
- 1Compare S3, EFS, FSx for NetApp ONTAP, RDS, and DynamoDB on cost, latency, and access pattern
- 2Load a CSV dataset into S3, convert it to Parquet with AWS Glue, and query it with Athena
- 3Build a Kinesis Data Streams to Firehose pipeline that lands records in S3
- 4Study when Parquet or ORC beats CSV or JSON for training data
- 5Practice 30 questions scoped to data ingestion and storage
- 1Run a SageMaker Data Wrangler flow that imputes missing values and removes outliers
- 2Create a feature group in SageMaker Feature Store and read it back for training
- 3Practice one-hot, binary, and label encoding and explain when each distorts a model
- 4Label a small image or text dataset with SageMaker Ground Truth
- 5Compare Glue DataBrew, Glue ETL, and Spark on EMR on operational effort
- 6Practice 30 questions on transformation, class imbalance, and data integrity
- 1List every SageMaker built-in algorithm with its problem type and input format
- 2Train XGBoost on a tabular dataset with SageMaker training jobs
- 3Fine-tune a JumpStart or Bedrock foundation model on a small custom dataset
- 4Run a SageMaker automatic model tuning job and read the tuning report
- 5Map each managed AI service to the business problem it solves cheapest
- 1Compute precision, recall, F1, and AUC by hand from a confusion matrix
- 2Decide which metric fits fraud detection, medical screening, and demand forecasting
- 3Run SageMaker Clarify on a trained model and read the bias and SHAP output
- 4Reproduce an overfitting curve and fix it with regularization and early stopping
- 5Use SageMaker Debugger to catch a vanishing gradient or stalled convergence
- 6Practice 30 questions on model evaluation and tuning
- 1Deploy the same model to a real-time endpoint, a serverless endpoint, and batch transform
- 2Record cold start, latency, and cost differences between those three options
- 3Configure an asynchronous endpoint and observe the S3 output and SNS notification
- 4Attach a target tracking auto scaling policy keyed on invocations per instance
- 5Compare GPU, CPU, and inference-optimized instance families for one workload
- 1Define a SageMaker endpoint and its IAM role in CloudFormation, then redeploy it with the AWS CDK
- 2Build a container image, push it to Amazon ECR, and serve it from SageMaker
- 3Chain CodeBuild, CodePipeline, and CodeDeploy into a model promotion pipeline
- 4Orchestrate a training and evaluation workflow with SageMaker Pipelines and again with Step Functions
- 5Practice explaining blue/green, canary, and linear rollouts with rollback triggers
- 1Enable SageMaker Model Monitor with a baseline and trigger a data quality violation
- 2Create CloudWatch alarms on model latency and endpoint invocation errors
- 3Build a CloudWatch dashboard covering endpoint, training job, and pipeline metrics
- 4Tag every ML resource and read the split in AWS Cost Explorer
- 5Compare Spot Instances, On-Demand, Reserved Instances, and SageMaker Savings Plans for training
- 1Write a least privilege IAM policy for a training job that reads one S3 prefix
- 2Place a SageMaker endpoint inside a VPC with no internet route and confirm it still serves
- 3Encrypt training data and model artifacts with a customer managed AWS KMS key
- 4Create a CloudTrail trail and locate the API call that started a training job
- 5Review Macie findings on a bucket that holds sample personal data
- 6Practice 30 questions on security, monitoring, and cost optimization
- 1Sit the 65-question Official Practice Exam under a strict 130-minute clock
- 2Score every domain separately and rank them from weakest to strongest
- 3Rework each missed question until you can explain why the distractors fail
- 4Sit a second timed practice exam from a different provider
- 5Re-read the exam guide task statements and confirm nothing is unfamiliar
- 6Book the exam and confirm your ID matches your AWS Certification Account name
Ready to pass AWS Certified Machine Learning Engineer - Associate (MLA-C01)?
Get 500+ practice questions, video walkthroughs, and a pass guarantee.
Best Study Resources
Exam Prep Plan: AWS Certified Machine Learning Engineer - Associate (MLA-C01)
Official learning planThe four-step AWS plan on Skill Builder with digital courses, flashcards, and more than 125 exam-style questions across its practice assessments.
Free, some content requires a Skill Builder subscription
AWS Certified Machine Learning Engineer - Associate (MLA-C01) Exam Guide
Official exam guideThe authoritative content outline with domain weightings, every task statement, and the in-scope and out-of-scope AWS service lists.
Free
Official Practice Question Set: MLA-C01
Official practice questionsTwenty AWS-written questions in the real formats, with explanations for every option. Use it early to calibrate rather than as a final readiness check.
Free
Official Practice Exam: MLA-C01
Official full-length mockSixty-five questions with a 130-minute limit that mirrors the live exam structure, including the unscored-item ratio and the four question formats.
Included with an AWS Skill Builder subscription at 29 USD per month or 449 USD per year
Amazon SageMaker AI Developer Guide
Vendor documentationThe reference for endpoint types, built-in algorithms, Feature Store, Model Monitor, Clarify, Pipelines, and Model Registry. Most exam disagreements resolve here.
Free
AWS Well-Architected Machine Learning Lens
WhitepaperDesign principles and best practices for ML workloads across the six pillars. It maps closely to the monitoring, maintenance, and security domain.
Free
Tutorials Dojo MLA-C01 Practice Exams
Third-party practice examsTimed and review-mode question banks with worked explanations and cheat sheets, sold with one year of access. A free sampler set is available first.
14.99 USD
AWS Certified Machine Learning Engineer Associate MLA-C01 on Udemy
Video courseThe Frank Kane and Stephane Maarek course covering SageMaker, Bedrock, and MLOps with hands-on demos. Useful for candidates without daily SageMaker exposure.
Varies with Udemy promotional pricing
AWS Certification policies: before, during, and after testing
Policy referenceThe source for ID rules, accommodations, regional pricing, break rules, the 14-day retake wait, and the five business day result window.
Free
AWS Machine Learning Blog
BlogImplementation write-ups from AWS engineers on SageMaker deployment patterns, drift detection, and cost tuning. Good for turning service names into concrete workflows.
Free
Common Mistakes to Avoid
Studying machine learning theory instead of AWS service selection
MLA-C01 rarely asks you to derive an algorithm. It asks which AWS service solves a stated problem at the lowest cost or operational effort. For every concept you learn, name the AWS service that implements it and the alternative you would reject.
Treating the four SageMaker inference options as interchangeable
Learn the decision rule. Real-time endpoints for steady low-latency traffic, serverless for spiky low-volume traffic that tolerates cold starts, asynchronous for large payloads or long inference times, and batch transform for scheduled scoring with no endpoint to maintain.
Skipping the in-scope service list because it looks like trivia
The list names roughly 70 services across analytics, containers, developer tools, and security. Question stems assume you know what Amazon Kendra, AWS Lake Formation, or Amazon Augmented AI do. Write one sentence per service and review it weekly.
Ignoring the 24 percent monitoring, maintenance, and security domain
That domain outweighs deployment at 22 percent. Spend real lab time on Model Monitor baselines, CloudWatch alarms, CloudTrail, IAM least privilege for training jobs, and cost allocation tagging, because those questions are answerable with hands-on memory.
Guessing on multiple response questions after finding one correct answer
AWS awards no partial credit. Evaluate all five or more options independently and confirm the count the question asks for. If you can only defend one of two required answers, flag the question and come back rather than locking in a half answer.
Practicing only on multiple choice banks
Ordering and matching items appear on the live exam and behave differently under time pressure. Use the Official Practice Question Set and Official Practice Exam, which include those formats, so exam day is not the first time you drag a five-step sequence into order.
Assuming SageMaker is always the right answer
Many questions are cheapest to solve with a managed AI service. Amazon Rekognition, Amazon Comprehend, Amazon Textract, Amazon Personalize, and Amazon Bedrock remove the training work entirely, and AWS often marks the custom SageMaker model as the wrong tradeoff.
Confusing SageMaker Clarify, Model Monitor, and Debugger
Clarify explains predictions and measures bias, Model Monitor watches deployed endpoints for data and model quality drift, and Debugger inspects training jobs for convergence problems. Run all three once so the distinction comes from memory rather than reasoning.
Leaving questions blank while chasing a perfect answer
Unanswered questions are scored as incorrect and there is no guessing penalty. Answer every question on the first pass, flag the uncertain ones, and use the review screen at the end to revisit only the flagged set.
Booking the exam without checking which version you will sit
Registration for MLA-C02 opens September 1, 2026, and September 28, 2026 is the last English MLA-C01 date. Confirm the exam code on your Pearson VUE confirmation email and study from the exam guide that matches that code.
Exam Day Tips
- 1
Budget two minutes per question. At 130 minutes for 65 questions there is no spare time for a long detour on one scenario.
- 2
Read the last sentence of the scenario first. It usually contains the constraint, such as lowest cost, least operational overhead, or lowest latency, that separates two technically correct answers.
- 3
Use the Help button when a short service name is unfamiliar. AWS publishes the short-name to full-name mapping inside the exam.
- 4
Flag rather than stall. The review screen at the end lets you filter to flagged and incomplete questions, so a first pass at speed is safer than perfect ordering.
- 5
Save exam feedback comments for the end. Time spent typing in the Comment box is subtracted from your exam time.
- 6
You have five minutes to read and accept the Candidate Code of Conduct. Timing out ends the exam with no refund.
- 7
Plan around breaks. Online proctored candidates may not leave the camera view for any reason, and unscheduled test center breaks keep the timer running plus force a full re-check-in.
- 8
Bring two forms of ID to a test center, with a government-issued photo ID as the primary. The name must match your AWS Certification Account.
- 9
For online proctoring, run the Pearson VUE system test on the same machine and network you will use, and clear the desk before check-in.
- 10
Watch for the five-minute warning in the upper right. Use it to convert every unanswered item into a best guess rather than a blank.
Career Paths & Salary Ranges
Machine learning engineer
Builds and ships models into production on SageMaker, owning the pipeline from feature store to endpoint. The Bureau of Labor Statistics projects 34 percent employment growth for data scientists from 2024 to 2034 and counted 245,900 jobs in 2024.
$112,590 median for data scientists (BLS, May 2024)
MLOps engineer
Automates training, evaluation, and deployment with SageMaker Pipelines, CodePipeline, and infrastructure as code. This is the role MLA-C01 maps to most directly, since domains 3 and 4 together carry 46 percent of the scored content.
$133,080 median for software developers (BLS, May 2024)
Data engineer for ML platforms
Owns ingestion, Glue transformations, and feature pipelines that supply model training. Domain 1 of MLA-C01 covers this work at 28 percent, and the AWS Certified Data Engineer - Associate exam is the natural pairing.
$123,100 median for database administrators and architects (BLS, May 2024)
ML security and governance engineer
Applies least privilege IAM, VPC isolation, KMS encryption, and CloudTrail auditing to ML workloads. Task 4.3 of the exam covers exactly this scope, including SageMaker Role Manager and securing CI/CD pipelines.
$124,910 median for information security analysts (BLS, May 2024)
Applied research engineer
Works on model architecture, fine-tuning, and evaluation rather than pipeline plumbing. BLS lists a master's degree as the typical entry-level education for this occupation, so the certification supplements rather than replaces graduate study.
$140,910 median for computer and information research scientists (BLS, May 2024)
Prerequisites & Requirements
- No formal prerequisites. AWS allows any candidate to register for MLA-C01 without holding another certification.
- AWS recommends at least one year of experience using Amazon SageMaker and other AWS services for ML engineering.
- AWS also recommends at least one year in a related role such as backend software developer, DevOps developer, data engineer, or data scientist.
- Working knowledge of common ML algorithms, data formats, and querying and transforming data.
- Experience with CI/CD pipelines, infrastructure as code, and version control with Git.
- Candidates aged 13 to 17 may test with the consent of a parent or legal guardian; there is no upper age limit.
- A valid government-issued ID establishing residence in a non-sanctioned country is required to sit the exam.
Frequently Asked Questions
How much does the AWS Certified Machine Learning Engineer - Associate exam cost?
One attempt costs 150 USD. AWS publishes local pricing of 128 EUR, 224 AUD, 20,000 JPY, 197,287 KRW, and 1,057 CNY, plus 12,829.50 INR through the Pearson Mindhub voucher store only. Applicable taxes may be added, and AWS updates the foreign exchange rates at least annually in May with a minimum of 30 days notice.
What happens if I fail MLA-C01?
You must wait 14 calendar days from the failed attempt before you can retake the exam. There is no limit on the number of attempts, but you pay the full registration fee each time. Your score report in the AWS Certification Account includes section-level performance classifications that show which domains pulled the score down.
How long do I have to wait to retake the exam?
Fourteen calendar days after a failed attempt. That wait applies to each failed attempt, not only the first. If you pass, AWS blocks you from retaking the same exam for two years, unless AWS republishes the exam under a new exam guide and series code, which makes the new version available to you.
What is the passing score and how is it calculated?
You need a scaled score of 720 out of a range of 100 to 1,000. AWS set the original standard with the modified Angoff technique, where a panel of subject matter experts rated each question against a minimally qualified candidate, then equated every later exam form to that standard. Scoring is compensatory, so only the total matters, not each domain.
How many questions do I need to answer correctly?
AWS does not publish a raw number of correct answers. Fifty of the 65 questions are scored, the other 15 are unscored trial items, and the raw cut point shifts slightly between exam forms because of statistical equating. Aiming to answer at least 40 of the 50 scored questions correctly gives you a reasonable margin.
When do I get my results?
Final results post to your AWS Certification Account within five business days of the exam, and AWS emails you when they are available. The Pearson VUE exit screen provides information about your results at the end of the appointment. Once posted, you can view, download, or print a PDF score report from the Exam History tab.
What is the pass rate for MLA-C01?
AWS does not publish pass rates for any certification exam, including MLA-C01. Neither the exam guide nor the certification FAQ contains a pass rate figure, so any percentage published elsewhere is a third-party estimate rather than an AWS statistic.
How long is the certification valid and how do I recertify?
The certification is valid for three years. AWS gives two renewal routes: pass the latest version of the AWS Certified Machine Learning Engineer - Associate exam, or pass the AWS Certified Generative AI Developer - Professional exam. Either route can use the 50 percent discount voucher in your AWS Certification Account, and either adds three years from the completion date.
What ID do I need to bring?
A valid government-issued photo ID establishing residence in a non-sanctioned country, plus a secondary form of ID. If you do not hold a qualifying government-issued ID from the country you are testing in, an international travel passport from your country of citizenship must be your primary ID, and a secondary ID is still required. The name on your ID must match your AWS Certification Account.
Can I take the exam online instead of at a test center?
Yes. Pearson VUE online proctoring is available for all AWS Certification exams, with English proctoring 24 hours a day, seven days a week. Japanese proctoring runs Monday to Saturday, and Spanish for Latin America and Mandarin for mainland China run Monday to Friday, each within set local-time windows. You need a private space, a webcam, and screen sharing, and you must speak with the proctor to complete check-in.
Can I take a break during the exam?
Not during an online proctored or kiosk appointment. AWS configures no scheduled breaks for any certification exam. At a test center you may take an unscheduled break, but the exam timer keeps running, you cannot leave the building, and you must complete the check-in steps again before you can resume.
What materials can I bring in with me?
None. AWS exams are closed book, and the Pearson VUE Candidate Rules Agreement governs test center conduct, with phones, notes, and smart devices stored outside the testing room. Specific comfort aids such as medicines and medical devices are permitted under the AWS accommodations policy. Inside the exam, the only reference is the Help button listing AWS service short names.
Are testing accommodations available?
Yes. AWS grants reasonable accommodations for documented disabilities through Pearson VUE, and accommodations must be requested before scheduling each exam. Separately, non-native English speakers taking the exam in English can request the ESL +30 accommodation for an extra 30 minutes. ESL +30 is requested once from your AWS Certification Account and then applies to all future registrations.
In which languages is the exam offered?
English, Japanese, Korean, and Simplified Chinese. Candidates registered for a translated version can view any question in English during the exam using the on-screen language toggle. Note that the updated MLA-C02 beta is English only at launch, with the other three languages arriving at general availability.
Can I reschedule or cancel my appointment?
Yes, up to 24 hours before the scheduled start time, with no additional fee. Cancelling more than 24 hours ahead refunds the fee you paid at purchase. AWS limits you to rescheduling an exam twice. Inside the 24-hour window you lose the fee, so rebook early if your plans change.
How does MLA-C01 compare to the AWS Certified AI Practitioner exam?
AI Practitioner (AIF-C01) is a foundational exam of 65 questions in 90 minutes for 100 USD, aimed at people who use AI services without building them. MLA-C01 is an associate engineering exam of 65 questions in 130 minutes for 150 USD that requires hands-on SageMaker work. Passing MLA-C01 also recertifies AIF-C01 for three years.
Should I take MLA-C01 or the AWS Certified Data Engineer - Associate exam first?
Take Data Engineer (DEA-C01) first if your daily work is ingestion, warehousing, and pipelines, since it shares the same 130-minute, 65-question, 150 USD format and overlaps with the 28 percent data preparation domain in MLA-C01. Take MLA-C01 first if you already train and deploy models and only need to formalize the MLOps side.
What replaced the AWS Certified Machine Learning - Specialty exam?
AWS retired ML Specialty (MLS-C01) after March 31, 2026, and existing holders keep an active credential for three years from the date they earned it. MLA-C01 now covers the associate-level ML engineering role, and the AWS Certified Generative AI Developer - Professional exam, a 180-minute, 75-question paper for 300 USD, covers the professional tier.
Should I wait for MLA-C02?
Register for MLA-C01 if you can sit before September 28, 2026, the last English date. Registration for MLA-C02 opens September 1, 2026, and the Korean, Japanese, and Simplified Chinese versions of MLA-C01 remain bookable until MLA-C02 reaches general availability. Both versions certify the same job role, and either earns a credential valid for three years.
How long should I study for MLA-C01?
Plan eight to twelve weeks at roughly ten hours per week if you meet the AWS recommendation of one year with SageMaker. Candidates without production SageMaker exposure usually need the longer end of that range, because domains 3 and 4 reward lab experience with endpoints, pipelines, Model Monitor, and IAM more than they reward reading.
Pass AWS Certified Machine Learning Engineer - Associate (MLA-C01) — Guaranteed
94% pass rate on first attempt
One-time • Lifetime access
One-time payment
AWS Certified Machine Learning Engineer - Associate (MLA-C01) Pass Rate
How hard is it? (Not published)
Failed AWS Certified Machine Learning Engineer - Associate (MLA-C01)?
Recovery plan & retake policy
Study Timeline
Week-by-week study plan
Requirements
Prerequisites & registration
Compare Certifications
Side-by-side comparisons