Start with the pipeline, not a list of services
Data engineering is the work of moving data from a source to a trustworthy consumer. A useful pipeline states what arrives, when it arrives, how it changes, what can fail and who may read the result. Before opening the AWS console, sketch those answers for one fictional dataset. SQL, basic Python control flow, relational concepts and a code editor will make the exercises easier; you do not need to master Spark or Airflow before the first week.
If you are new to AWS, begin with account safety, IAM roles and policies, CLI profiles and a cost plan. Avoid daily work through the root user. Use synthetic data so your portfolio can be shared without exposing a customer record. The goal of each week below is a reviewable artifact—an explanation, diagram, test result or runbook—not simply a service you clicked.
AWS calls DEA-C01 an Associate exam for candidates with substantial data-engineering and AWS experience. This roadmap can help build that experience, but the official AWS exam guide defines exam scope and should be checked before booking.
A flexible 8–10 week learning sequence
Use the periods as milestones, not fixed class dates. If SQL or AWS basics are new, give them more time; if you already operate pipelines, spend more time on failure recovery, governance and measurable performance. Optional weeks 9–10 are for integration and review, not a claim that every candidate will be exam-ready within ten weeks.
Foundations and a safe sandbox
Refresh SQL and Python control flow; set up a separate learning account or authorized sandbox, IAM role, CLI profile and cost alerts. Sketch a source-to-consumer data flow before choosing services.
Deliverable: A one-page source contract, cost checklist and least-privilege account plan.
S3 and event-driven basics
Organize a private raw-data area, upload small synthetic CSV and JSON files, compare prefixes and partitions, and test versioning. Add one S3 event that invokes Lambda, with output filtered to a different prefix or bucket.
Deliverable: A documented object layout and an invocation test that cannot recurse on its own output.
Catalog and batch transformation
Use Glue Data Catalog and a crawler to inspect a sample schema. Define validation rules before a Glue transformation writes curated data. Test a malformed row, a second run and a controlled backfill.
Deliverable: A batch pipeline diagram plus accepted, rejected and replay counts.
Athena and lake permissions
Query curated S3 data in Athena. Compare scanned bytes before and after a partition or file-format decision. Trace IAM, S3, Lake Formation and KMS permissions for one reporting role.
Deliverable: A query-cost comparison and an access trace with an intentionally denied read.
Operational and analytical stores
Compare an RDS source with a Redshift reporting destination. Try a small S3-to-Redshift COPY design and explain why warehouse loading, transactional writes and Athena queries solve different problems.
Deliverable: A store-selection table and an ingestion plan with load validation.
Streaming and larger processing
Model Kinesis record arrival, partition keys and late data. Decide whether Glue, EMR on EC2 or EMR Serverless is appropriate for a Spark transformation; do not start a cluster merely to learn service names.
Deliverable: A batch-versus-streaming decision and a resource-cleanup checklist.
Messages, secrets and deployment
Use SNS and SQS in a failure-notification design, store sample credentials in Secrets Manager, and express a small repeatable stack with CloudFormation or CDK. Explain what each component owns.
Deliverable: An infrastructure diagram and a deployment/rollback note without real secrets.
Operate and explain the pipeline
Add CloudWatch signals for source lag, failed jobs and data quality; review spend with AWS Budgets. Compare Glue Workflows, Step Functions and Airflow-style orchestration by the control you need, not by popularity.
Deliverable: A runbook with alert owner, retry limit, replay steps and a demo of one failed stage.
Portfolio and targeted review
Combine the strongest batch and streaming exercises into a small portfolio. Document trade-offs, review the current DEA-C01 guide by domain and revisit weak decisions with original practice questions.
Deliverable: Architecture diagrams, safe sample data, README, measured results and a domain-tagged mistake log.
Make service choices from the workload
S3, Glue, Athena and Lake Formation
S3 stores objects; Glue catalogs and transforms; Athena queries data in S3; Lake Formation helps govern supported lake resources. A catalog entry is not a permission grant or proof of data quality. Partition on fields common queries filter by, then measure actual bytes scanned rather than assuming every partition layout saves money.
RDS and Redshift
RDS is a managed relational database that can be a source for operational records. Redshift serves analytical warehousing. Before loading, define how updates, duplicate keys and failed batches are reconciled. A simple S3 staging and COPY exercise can reveal the difference between successful file transfer and correct reporting data.
Kinesis, Glue and EMR
Choose a streaming path only when arrival latency and event volume justify it. Glue can support managed ETL; EMR on EC2 exposes primary, core and task nodes for cluster-based processing, while EMR Serverless runs supported jobs without managing a cluster. Compare operations, scale and cost against the workload before launching resources.
CloudWatch and orchestration
CloudWatch can surface job failures, lag and quality signals; AWS Budgets helps track spend. Glue Workflows, Step Functions and Airflow-style DAGs express different kinds of coordination; first write the dependency, retry and failure-handling requirements. SNS or SQS may help notify or decouple steps, and Secrets Manager protects credentials that should not sit in code.
Three portfolio projects that demonstrate decisions
Choose one project first and make it reliable before building the others. For every project, record the source contract, access boundary, data-quality rule, replay plan and expected cost. Publish diagrams and synthetic samples rather than real account secrets or sensitive data.
Governed batch analytics
Design: Synthetic orders → private S3 raw area → Glue catalog and validation → curated partitioned files → Athena query.
Prove it: Show accepted/rejected counts, an event-date predicate, scanned-byte comparison and a denied query for a role without permission.
Warehouse reporting
Design: Small relational or file source → documented transformation → S3 staging → Redshift COPY → reporting table.
Prove it: Explain store choice, load idempotency, row reconciliation, failure alert and the cost of leaving compute running.
Streaming with a recovery path
Design: Kinesis sample events → transformation → S3 analytical output → monitoring and replay plan.
Prove it: Test duplicate IDs, late events and a failed downstream stage; demonstrate how output is recovered without silently double-counting.
Keep your README honest: these are ITCertPath-suggested exercises, not claims that the linked introductory video demonstrates completed projects or that a portfolio proves a certification result.
Connect the skills to DEA-C01, then check readiness separately
The current AWS Certified Data Engineer – Associate exam code is DEA-C01. Its four scored domains are data ingestion and transformation (34%), data store management (26%), data operations and support (22%), and data security and governance (18%). Use the official guide to map what you built to each domain; the service-focused schedule above is not a complete exam blueprint.
- Explain one batch and one near-real-time ingestion choice, including a failure and replay path.
- Defend a store and partition decision using the consuming query and measured trade-offs.
- Show how you detect bad data, alert an owner and recover without double-counting.
- Trace one principal through IAM, data-lake permissions and encryption controls.
Next, use the existing ITCertPath DEA-C01 course outline for domain-level learning units and explained exercises. The original DEA-C01 practice series helps identify weak decisions; its scores are not AWS scaled exam scores. Review the AWS voucher and booking guidance only after checking the current exam identity, country, price and validity with AWS.
Keep every hands-on lab safe and affordable
- Use an authorized account and small synthetic data. Check current service pricing, set cost alerts, monitor usage and remove EC2, EMR, Redshift and other resources after each lab. An alert is not a hard spend limit.
- Keep S3 event-triggered Lambda output in a separate bucket or filtered prefix so the function does not repeatedly trigger itself. Test the filter before a larger upload.
- Use roles and narrow permissions; do not publish access keys, real customer data or Secrets Manager values in a GitHub portfolio.
- Write the rollback and cleanup instructions alongside the deployment steps. An architecture that cannot be recovered is not portfolio-ready.
Frequently asked questions
Can a beginner use this AWS data-engineering roadmap?
Yes, as a learning sequence. Begin with SQL, Python basics and cloud foundations, and use small safe exercises. AWS describes the DEA-C01 target candidate as having 2–3 years of data-engineering experience and 1–2 years of hands-on AWS experience, so completing this sequence alone does not establish exam readiness.
Does this 8–10 week plan describe an ITCertPath live batch?
No. It is a suggested self-paced study sequence, not an advertised class duration, confirmed instructor-led batch or guarantee of certification. Ask ITCertPath separately for current training format and availability.
Is this the complete DEA-C01 exam syllabus?
No. This practical roadmap covers useful service and pipeline decisions, but AWS's current official DEA-C01 guide defines the exam domains and can change. Compare every preparation decision with that guide before booking.
Do I need to launch EC2, EMR and Redshift to follow the guide?
No. You can first design and explain the decisions with diagrams and sample data. If you run optional labs, use an authorized account, review pricing, set alerts and clean up resources. Budget alerts do not impose a hard spending cap.
Sources and editorial note
This is an independently written ITCertPath study guide. A reader shared an introductory curriculum video as a starting point; the sequence, safeguards and exercises here are our own editorial organization. We have not copied exam questions, verified delivery of that video's proposed labs, or claimed AWS or the creator endorses this page.
Provider documentation changes. Last editorial check: 24 October 2026. How ITCertPath reviews and corrects guides.