MLA-C01 — AWS Certified Machine Learning Engineer – Associate Study Plan
A practical 7-day, 14-day, 30-day, and 60/90-day study plan for the AWS Certified Machine Learning Engineer – Associate (MLA-C01) exam.
Who this study plan is for
This Study Plan is for candidates preparing for the AWS Certified Machine Learning Engineer – Associate (MLA-C01) exam from AWS. It is designed for practical preparation: diagnostic practice, AWS service review, machine learning workflow drills, missed-question analysis, timed mock exams, and final-week review.
Use the plan that matches your time remaining. If you are unsure where to start, take a short diagnostic first and let your misses decide the order of review.
Which plan should you use?
| Time until exam | Best fit | Primary goal | Main risk to manage |
|---|---|---|---|
| 7 days | Final review plan | Stabilize weak areas and build exam timing | Trying to learn too much new material |
| 14 days | Focused plan | Cover high-value gaps with daily practice | Skipping missed-question review |
| 30 days | Balanced plan | Combine domain review, AWS service practice, and timed exams | Studying services without scenario practice |
| 60/90 days | Full preparation path | Build ML engineering depth and repeatable exam readiness | Moving too slowly or avoiding timed practice |
If you have hands-on AWS experience but limited exam prep, start with the 14-day or 30-day plan. If you are newer to AWS ML workflows, use the 60/90-day path.
Core MLA-C01 study areas to organize around
Use the current AWS exam guide as your authority, then organize study time around these practical workstreams.
| Study area | What to know how to reason through |
|---|---|
| Data preparation | Data sources, ingestion, storage, feature engineering, data quality, labeling, transformation, partitioning, and data leakage prevention |
| Model development | Algorithm and framework selection, training jobs, evaluation metrics, tuning, overfitting, underfitting, bias/variance, and experiment tracking |
| ML operations | Pipelines, orchestration, model registry, deployment patterns, rollback, retraining, automation, and CI/CD concepts |
| Monitoring and troubleshooting | Logs, metrics, drift, data quality changes, endpoint performance, failed training jobs, cost symptoms, and operational alarms |
| Security and governance | IAM permissions, least privilege, encryption, network isolation, S3 access, KMS, VPC endpoints, audit logging, and responsible data handling |
| AWS service selection | When to use SageMaker capabilities, S3, Glue, Athena, Lambda, Step Functions, EventBridge, CloudWatch, CloudTrail, ECR, CodePipeline, and related AWS services |
Daily practice rhythm
Use this rhythm on most study days, regardless of plan length.
| Block | Time | What to do |
|---|---|---|
| Warm-up recall | 10 minutes | Write down key services, metrics, or failure modes from memory before reading notes |
| Targeted concept review | 30-45 minutes | Study one topic: data prep, training, deployment, monitoring, security, or troubleshooting |
| Scenario practice | 30-45 minutes | Answer practice questions or walk through AWS architecture decisions |
| Missed-question review | 20-30 minutes | Classify each miss by cause and write a correction note |
| Hands-on or diagram review | 20-40 minutes | Review a SageMaker workflow, pipeline, IAM policy shape, deployment path, or monitoring setup |
| Closeout | 5 minutes | Pick tomorrow’s weakest topic based on today’s evidence |
For shorter timelines, reduce reading time before reducing missed-question review. Reviewing why you missed questions is usually more valuable than adding another passive lesson.
Diagnostic-first setup
Before starting a 14-day, 30-day, or 60/90-day path, complete a diagnostic session.
| Step | Action |
|---|---|
| 1 | Take a mixed set of practice questions without notes |
| 2 | Tag every miss by topic and cause |
| 3 | Build a weak-area list with no more than 5 priorities |
| 4 | Schedule those priorities first, not last |
| 5 | Retest the same topics after review using new questions |
Use these miss categories:
| Miss type | Meaning | Fix |
|---|---|---|
| Concept gap | You did not know the ML or AWS concept | Study the concept, then answer targeted questions |
| Service-selection gap | You knew the topic but chose the wrong AWS service or feature | Compare service use cases side by side |
| Scenario-reading error | You missed constraints such as latency, security, scale, or automation | Underline constraints before choosing an answer |
| Troubleshooting gap | You could not identify the likely cause or next step | Build a symptom-to-cause table |
| Overthinking | You rejected the simple AWS-native answer without a reason | Practice choosing the best fit, not the most complex design |
7-day final review plan
Use this if your exam is in one week. The goal is not to relearn everything. The goal is to remove avoidable mistakes, sharpen AWS service decisions, and build timing confidence.
| Day | Focus | Study actions |
|---|---|---|
| 1 | Diagnostic and triage | Take a timed mixed quiz or short mock. Build a weak-area list. Review the current AWS exam guide. Identify your top 4 gaps. |
| 2 | Data preparation and feature workflows | Review S3 data layout, Glue/Athena use cases, data quality, feature engineering, labeling, leakage, train/validation/test splits, and preprocessing choices. |
| 3 | Model training and evaluation | Review SageMaker training concepts, algorithm/framework choice, hyperparameter tuning, evaluation metrics, overfitting, underfitting, and experiment comparison. |
| 4 | Deployment and MLOps | Review endpoints, batch inference, model registry, pipelines, orchestration, container images, rollback, retraining triggers, and automation patterns. |
| 5 | Monitoring, troubleshooting, and security | Review CloudWatch, CloudTrail, logs, drift, endpoint errors, failed jobs, IAM, KMS, S3 access, VPC/network controls, and least privilege. |
| 6 | Timed mock and deep review | Take a timed mock exam or the longest realistic timed set available. Spend at least as long reviewing as testing. Create a final correction sheet. |
| 7 | Light final review | Review correction sheet, key service comparisons, metrics, and common traps. Do not add large new topics. Prepare exam-day logistics. |
7-day rule
Stop adding new major material after Day 5. On Days 6-7, focus on:
- missed-question explanations
- service-selection comparisons
- common troubleshooting scenarios
- security and monitoring reminders
- timing and confidence
14-day focused plan
Use this if you have two weeks and can study most days. This plan assumes you already have some AWS or ML background and need focused exam readiness.
| Day | Focus | Study actions |
|---|---|---|
| 1 | Diagnostic | Take a mixed diagnostic. Tag every miss. Build a 5-topic weak-area list. |
| 2 | Exam guide mapping | Map weak areas to the AWS exam guide. Set up a study tracker by topic. |
| 3 | Data ingestion and storage | Review S3, data formats, data access, Glue, Athena, data catalog concepts, and secure access patterns. |
| 4 | Data preparation | Drill data cleaning, labeling, feature engineering, leakage prevention, imbalance, and transformation choices. |
| 5 | Model training | Review SageMaker training jobs, built-in algorithms versus custom training, containers, compute selection concepts, and training artifacts. |
| 6 | Evaluation and tuning | Drill metrics, validation strategy, tuning, model comparison, overfitting, underfitting, and threshold tradeoffs. |
| 7 | Mixed practice checkpoint | Take a timed mixed set. Review misses. Update weak-area list. |
| 8 | Deployment patterns | Review real-time endpoints, batch inference, model packaging, rollout, rollback, and latency/cost tradeoffs. |
| 9 | MLOps and orchestration | Review SageMaker Pipelines, Step Functions, EventBridge, Lambda, CI/CD concepts, model registry, and retraining workflows. |
| 10 | Monitoring and troubleshooting | Drill CloudWatch, logs, metrics, drift, data quality changes, endpoint failures, job failures, and alerting patterns. |
| 11 | Security and governance | Review IAM, least privilege, S3 policies, KMS, VPC isolation, CloudTrail, auditability, and sensitive data controls. |
| 12 | Timed mock exam | Take a full-length or near full-length timed mock. Do not pause. Simulate exam conditions. |
| 13 | Weak-area sprint | Rework every missed mock question. Drill only the weakest 3-4 topics. Build final notes. |
| 14 | Final review | Light review only. Revisit correction sheet, service comparisons, and exam-day timing plan. |
Best use of the 14-day plan
Spend at least 40% of your total study time on practice and review, not reading. The exam rewards applying AWS ML engineering concepts to scenarios.
30-day balanced plan
Use this if you want a realistic preparation window with time for concept review, hands-on reinforcement, practice questions, and multiple timed checkpoints.
30-day weekly structure
| Week | Main objective | Output by end of week |
|---|---|---|
| Week 1 | Baseline and data preparation | Diagnostic complete, data workflow gaps identified, first missed-question log built |
| Week 2 | Model development and evaluation | Training, tuning, metrics, and algorithm-selection notes completed |
| Week 3 | Deployment, MLOps, monitoring, and security | Deployment and operations scenarios drilled |
| Week 4 | Timed exams and weak-area closure | Mock exams reviewed, final correction sheet ready |
Days 1-7: Baseline and data workflows
| Day | Focus | Study actions |
|---|---|---|
| 1 | Diagnostic | Take a mixed diagnostic. Build your topic tracker. |
| 2 | AWS ML workflow overview | Diagram an end-to-end ML workflow: data source, storage, processing, training, registry, deployment, monitoring, retraining. |
| 3 | Data storage and access | Review S3, data formats, access control, encryption concepts, and query/discovery services such as Glue and Athena. |
| 4 | Data preparation | Study preprocessing, cleaning, labeling, feature engineering, class imbalance, and leakage prevention. |
| 5 | Feature and pipeline thinking | Review repeatable transformations, feature reuse, pipeline steps, and training/serving consistency. |
| 6 | Data scenario practice | Complete targeted practice on data preparation and service selection. |
| 7 | Review day | Rework all missed data questions. Make a one-page data-prep correction sheet. |
Days 8-14: Model development
| Day | Focus | Study actions |
|---|---|---|
| 8 | Training workflows | Review SageMaker training jobs, artifacts, containers, input/output data, and job failure signals. |
| 9 | Algorithms and frameworks | Compare built-in algorithms, custom training, managed notebooks, and framework choices at a high level. |
| 10 | Evaluation metrics | Drill classification, regression, ranking, and forecasting metric selection where relevant. |
| 11 | Tuning and optimization | Review hyperparameter tuning, validation strategy, early stopping concepts, overfitting, and underfitting. |
| 12 | Experiment comparison | Practice scenarios that require choosing the better model based on metrics and business constraints. |
| 13 | Model development quiz | Take a timed targeted set on training and evaluation. |
| 14 | Review day | Update missed-question log and rewrite weak concepts in your own words. |
Days 15-21: Deployment, MLOps, monitoring, and security
| Day | Focus | Study actions |
|---|---|---|
| 15 | Deployment patterns | Review real-time inference, batch inference, async-style decision factors, latency, throughput, and cost tradeoffs. |
| 16 | Model registry and release | Study model approval, packaging, versioning, rollback, and promotion between environments. |
| 17 | Orchestration | Review SageMaker Pipelines, Step Functions, Lambda, EventBridge, and scheduled or event-driven workflows. |
| 18 | CI/CD and infrastructure concepts | Review how code, containers, model artifacts, and infrastructure changes move through controlled release paths. |
| 19 | Monitoring | Study CloudWatch metrics/logs, model/data drift concepts, data quality, endpoint health, and alerting. |
| 20 | Security | Review IAM, least privilege, KMS, S3 access, VPC/network isolation, CloudTrail, and audit patterns. |
| 21 | Mixed operations practice | Take a timed mixed set focused on deployment, MLOps, monitoring, troubleshooting, and security. |
Days 22-30: Timed readiness and final weak-area sprint
| Day | Focus | Study actions |
|---|---|---|
| 22 | Mock exam 1 | Take a timed mock or long timed set. Simulate exam conditions. |
| 23 | Mock review | Review every question, including correct guesses. Tag misses by cause. |
| 24 | Weak area 1 | Deep review your lowest-scoring topic. Do targeted practice. |
| 25 | Weak area 2 | Deep review your second-lowest topic. Create service-selection notes. |
| 26 | Troubleshooting sprint | Drill failed training jobs, bad metrics, endpoint errors, permissions issues, data drift, and cost/performance symptoms. |
| 27 | Security and governance sprint | Review IAM, encryption, network controls, logging, auditability, and data handling scenarios. |
| 28 | Mock exam 2 | Take another timed mock or long timed set. Use strict timing. |
| 29 | Final correction sheet | Review misses. Create a concise final sheet: services, metrics, deployment patterns, security reminders, common traps. |
| 30 | Light final review | No heavy new material. Review final sheet, rest, and prepare exam logistics. |
60/90-day full preparation path
Use this if you are building stronger AWS ML engineering foundations or studying around a full-time job. The 60-day version compresses the checkpoints. The 90-day version adds more hands-on reinforcement and repetition.
60/90-day phase map
| Phase | 60-day timing | 90-day timing | Goal |
|---|---|---|---|
| Phase 1: Foundations and diagnostic | Days 1-7 | Weeks 1-2 | Establish baseline and understand exam scope |
| Phase 2: Data engineering for ML | Days 8-18 | Weeks 3-4 | Build confidence in data prep, access, quality, and features |
| Phase 3: Model development | Days 19-30 | Weeks 5-6 | Practice training, metrics, tuning, and model selection |
| Phase 4: Deployment and MLOps | Days 31-42 | Weeks 7-8 | Review production ML workflows and automation |
| Phase 5: Monitoring, troubleshooting, security | Days 43-50 | Weeks 9-10 | Build operational and governance readiness |
| Phase 6: Timed practice and final review | Days 51-60 | Weeks 11-13 | Convert knowledge into exam performance |
Phase 1: Foundations and diagnostic
| Task | What to do |
|---|---|
| Read the exam guide | Use the official AWS exam guide to define your topic list. Do not rely only on course chapter titles. |
| Take a diagnostic | Complete a mixed untimed or lightly timed question set. |
| Build a tracker | Track topic, confidence, practice score, and last review date. |
| Diagram the workflow | Draw the full ML lifecycle on AWS from raw data through monitoring and retraining. |
| Identify hands-on gaps | Mark services you have only read about but never used or diagrammed. |
Phase 2: Data engineering for ML
| Topic | Review actions |
|---|---|
| Data sources and ingestion | Compare batch, streaming, database, file, and event-driven data movement at a scenario level. |
| Storage and discovery | Review S3 organization, data catalog concepts, Glue, Athena, and secure data access. |
| Data quality | Practice identifying missing values, schema changes, skew, outliers, duplication, and leakage risks. |
| Feature engineering | Review transformation, encoding, scaling, aggregation, feature reuse, and training/serving consistency. |
| Labeling and supervised learning prep | Know how labeling quality affects model performance and evaluation. |
| Security | Include IAM, encryption, access boundaries, and audit logging in every data design review. |
Suggested weekly output:
- one data workflow diagram
- one service-selection comparison sheet
- one missed-question log update
- one targeted timed quiz
Phase 3: Model development
| Topic | Review actions |
|---|---|
| Training jobs | Review inputs, outputs, artifacts, logs, failure symptoms, and managed training concepts. |
| Algorithms and frameworks | Practice choosing between built-in capabilities, custom code, and framework-based training. |
| Evaluation metrics | Know which metric fits the business goal and data shape. |
| Tuning | Review hyperparameter tuning, validation strategy, overfitting, underfitting, and generalization. |
| Experiment management | Compare model versions using evidence, not intuition. |
| Cost/performance thinking | Understand how training choices affect runtime, repeatability, and operational complexity without memorizing unsupported limits. |
Phase 4: Deployment and MLOps
| Topic | Review actions |
|---|---|
| Inference patterns | Compare real-time, batch, and asynchronous-style workload needs by latency, volume, and cost constraints. |
| Model packaging | Review containers, artifacts, dependencies, and environment consistency. |
| Release workflow | Review model registry, approval, promotion, versioning, rollback, and controlled deployment. |
| Orchestration | Compare SageMaker Pipelines, Step Functions, Lambda, EventBridge, and scheduled workflows. |
| CI/CD concepts | Understand how model code, infrastructure, and artifacts are tested and promoted. |
| Retraining | Practice scenarios with drift, new data, scheduled retraining, and event-triggered retraining. |
Phase 5: Monitoring, troubleshooting, and security
| Topic | Review actions |
|---|---|
| Monitoring | Review endpoint health, latency, errors, logs, metrics, alarms, and model/data quality signals. |
| Drift and degradation | Practice identifying whether symptoms point to data drift, concept drift, bad features, bad labels, or infrastructure issues. |
| Troubleshooting | Drill permission failures, failed jobs, container errors, missing artifacts, bad input format, and unexpected model behavior. |
| Security | Review IAM least privilege, KMS encryption, S3 controls, VPC/network isolation, CloudTrail, and auditability. |
| Governance | Practice choosing controls that protect data and make ML workflows repeatable and traceable. |
Phase 6: Timed practice and final review
| Timing | Action |
|---|---|
| 10-14 days out | Take first full timed mock or long timed mixed set. |
| 7-10 days out | Review weak topics and retake targeted sets. |
| 5-7 days out | Take second timed mock or long timed set. |
| 3-4 days out | Stop adding major new topics. Review final correction sheet. |
| 1-2 days out | Light review only. Focus on timing, service comparisons, and common traps. |
Hands-on review ideas for MLA-C01
You do not need to build a large production system to prepare, but you should be able to reason through real AWS ML workflows. Use small, controlled exercises or architecture walkthroughs.
| Hands-on or diagram task | What to learn from it |
|---|---|
| Diagram a SageMaker training workflow | Where data enters, where artifacts go, where logs and metrics appear |
| Compare batch and real-time inference | Latency, cost, operational overhead, and scaling considerations |
| Trace an IAM permission failure | How roles, policies, S3 access, KMS, and service permissions interact |
| Design a retraining pipeline | Trigger, preprocessing, training, evaluation, approval, deployment, and monitoring |
| Review CloudWatch signals | Which metrics or logs help troubleshoot failed jobs and endpoint issues |
| Build a model registry flow | How versions move from experiment to approved deployment |
| Create a data quality checklist | Schema, missing values, distribution shift, labels, leakage, and feature consistency |
AWS service-selection review table
Use this table as a study aid. Do not memorize it as a replacement for scenario practice.
| Scenario clue | Services or concepts to review | Decision question |
|---|---|---|
| Raw data in object storage | Amazon S3, IAM, KMS, Glue, Athena | How is data organized, secured, discovered, and read by ML jobs? |
| Repeatable preprocessing | SageMaker processing concepts, Glue, pipelines | Should the transformation be reusable and automated? |
| Managed training | Amazon SageMaker training, containers, algorithms/frameworks | Is the training workflow managed, repeatable, and observable? |
| Model tuning | Hyperparameter tuning, validation strategy, metrics | What objective metric is being optimized? |
| Real-time prediction | SageMaker endpoint concepts, monitoring, autoscaling concepts | What latency and availability constraints matter? |
| Batch scoring | Batch inference patterns, S3 input/output | Is the workload offline or scheduled? |
| Workflow orchestration | SageMaker Pipelines, Step Functions, EventBridge, Lambda | Is this an ML pipeline, a general workflow, or an event trigger? |
| Monitoring | CloudWatch, logs, model/data monitoring concepts | What signal proves the system is healthy or degrading? |
| Audit and governance | CloudTrail, IAM, KMS, model registry, approvals | Can changes and access be traced? |
| Private access and isolation | VPC design concepts, security groups, endpoints, IAM | Does the workload need network isolation or private service access? |
Missed-question review method
A missed-question log is one of the highest-value tools for MLA-C01 preparation.
| Field | What to write |
|---|---|
| Question topic | Data prep, training, deployment, monitoring, security, troubleshooting, or service selection |
| Why you missed it | Concept gap, service confusion, misread scenario, timing, or overthinking |
| Correct reasoning | The minimum explanation that proves why the right answer fits |
| Wrong-answer trap | Why the tempting answer is not best |
| Follow-up action | Read notes, compare services, do hands-on review, or answer more targeted questions |
| Retest date | When you will test the topic again |
Example correction format
Use short entries. Long notes are harder to review in the final week.
| Prompt | Correction |
|---|---|
| What was the scenario asking for? | Secure, repeatable model deployment with approval and rollback |
| What did I choose incorrectly? | A manual deployment path |
| What should I remember? | Prefer controlled release, versioning, registry/approval, and rollback when governance is central |
| What will I drill next? | Deployment and MLOps scenarios |
When to use timed mock exams
Timed mocks are most useful after you have enough coverage to learn from the result.
| Timeframe | Mock strategy |
|---|---|
| 60/90-day plan | First long timed set around the midpoint, then 2-3 more during final phase |
| 30-day plan | One mock around Day 22 and another around Day 28 |
| 14-day plan | One full or near full timed mock around Day 12 |
| 7-day plan | One timed mock or long timed set on Day 6, earlier only if you need urgent triage |
How to review a mock
Do not only check the score. Review in this order:
- Questions you missed.
- Questions you guessed correctly.
- Questions that took too long.
- Questions where two answers looked correct.
- Topics that appeared repeatedly.
- Any scenario clues you ignored.
Create a final weak-area list with no more than 5 topics. If the list is longer, group similar issues together.
Scenario-reading checklist
MLA-C01 questions often reward identifying constraints before choosing a service or design. Use this checklist while practicing.
| Constraint | Ask yourself |
|---|---|
| Latency | Is prediction needed immediately or can it run later? |
| Scale | Is the workload small, bursty, recurring, or high-volume? |
| Automation | Is this a one-time task or a repeatable pipeline? |
| Security | Are there requirements for encryption, least privilege, private access, or audit logs? |
| Data quality | Is the problem caused by bad input data, drift, missing values, or inconsistent features? |
| Model quality | Is the issue metric choice, underfitting, overfitting, bias, or poor labels? |
| Operations | Is the system failing during training, deployment, inference, monitoring, or retraining? |
| Cost | Is the best answer the one that avoids unnecessary always-on resources or manual overhead? |
Final-week rules
Use these rules during the last 7 days, even if you followed a longer plan.
| Rule | Why it matters |
|---|---|
| Stop adding major new topics 3-4 days before the exam | New material can displace high-value review |
| Review misses more than notes | Your misses show where exam points are leaking |
| Keep practice timed | Timing problems are hard to fix without timed work |
| Revisit security daily | IAM, encryption, logging, and access control appear across many scenario types |
| Practice service selection | Many questions test choosing the best AWS-native path for a constraint |
| Sleep before the exam | Fatigue increases misreading and overthinking |
Exam-readiness checks
You are closer to ready when you can do the following without heavy notes.
| Readiness check | Can you do it? |
|---|---|
| Explain an end-to-end AWS ML workflow from data to monitoring | Yes / No |
| Choose between batch and real-time inference based on scenario constraints | Yes / No |
| Identify likely causes of failed training, poor model quality, or endpoint issues | Yes / No |
| Match evaluation metrics to model goals and data types | Yes / No |
| Reason through IAM, S3, KMS, VPC, and logging controls in ML scenarios | Yes / No |
| Explain when orchestration, pipelines, registry, approval, and rollback matter | Yes / No |
| Complete timed mixed practice without rushing the final questions | Yes / No |
| Review missed questions and avoid repeating the same mistake type | Yes / No |
If several answers are “No,” do not just take another mock. Spend a focused session on the weakest topics, then retest with targeted questions.
Practical next step
Choose your timeline, take a diagnostic practice set, and build a missed-question log today. Then follow the matching schedule above and adjust each day based on evidence from practice, not on how familiar a topic feels.