AWS Builder Center
How I cleared the AWS Solutions Architect Professional exam in 16 days using AI

How I cleared the AWS Solutions Architect Professional exam in 16 days using AI

Roger Nem (AWS Technical Account Manager) shares a four-round, AI-assisted study framework using Kiro that compressed preparation for the notoriously difficult AWS Solutions Architect Professional (SAP-C02) exam into 16 focused days

First and foremost, I give glory to Jesus for blessing me in passing one of the most demanding certification exams. I am also very thankful for my family for their patience and support during the time of preparation.🙏

Table of contents

  1. The challenge
  1. Exam domains and weightings
  1. My background
  1. A different approach
  1. My timeline
  1. Round 1: Keyword identification
  1. Round 2: Elimination technique
  1. Round 3: Full practice with explanations
  1. Round 4: Note consolidation
  1. Exam day strategy
  1. Key services to know deeply
  1. What worked and what I would do differently
  1. Exam day & results
  1. Final thoughts

The challenge

The  AWS Certified Solutions Architect — Professional (SAP-C02)  exam is widely regarded as one of the most challenging certifications in the cloud industry. It validates advanced technical skills in designing optimized AWS solutions based on the AWS Well-Architected Framework, spanning multi-account architectures, migrations, cost optimization, and security at scale.
What makes this exam particularly demanding is not just the breadth of services covered. It is the format itself. The questions are scenario-based, often consisting of one or two dense paragraphs packed with intentionally verbose context designed to test your ability to extract relevant information under time pressure. Each answer option is equally detailed, requiring you to evaluate multiple architectural trade-offs before selecting the most appropriate solution.
Fair warning: This exam is brutal. The questions and answer options are longer than any certification exam I have encountered. Amazon allocates 180 minutes for 75 questions, which works out to approximately 2 minutes and 24 seconds per question. That sounds reasonable until you realize each question can take 30–60 seconds just to read through the scenario before you even begin evaluating the options.
AWS Certified Solutions Architect - Professional (SAP-C02) Exam format and pricing as of the time of writing


Exam domains and weightings

The SAP-C02 exam evaluates proficiency across four domains:
AWS Certified Solutions Architect — Professional (SAP-C02) exam domains and weightings (picture by author)

Of the 75 questions, 65 are scored and 10 are unscored (experimental questions used by AWS for future exam development). You will not know which questions are unscored, so treat every question with equal attention.
This certification is valid for 3 years, after which you must recertify by taking the latest version of the exam.

My background

I am an Enterprise Technical Account Manager (TAM) at  Amazon Web Services (AWS)  with over 16 years of technology leadership experience. I have successfully guided global enterprise architecture, application development, and IT operations initiatives across 200+ markets worldwide. I am also a Security Technical Field Community specialist, helping enterprise customers design secure cloud architectures aligned with industry best practices.
Prior to joining AWS, I served as Principal Solutions Architect for Fortune 500 companies, where I drove strategic technical solutions at scale. My background combines deep technical expertise with strategic business acumen to help organizations across multiple industries accelerate their digital transformation initiatives and maximize their cloud investments.
I hold AWS certifications, including the AWS Certified Solutions Architect - Associate and AWS Certified Security - Specialty, which I cleared in 12 days, and several AWS Technology and Industry Accreditation badges —  https://www.credly.com/users/roger-nem/badges#credly .
I work with AWS services daily in my professional role, which gave me a solid foundation. However, the SA Professional exam covers a significantly broader surface area than any specialty exam, touching on networking, databases, analytics, migrations, containers, serverless, and organizational governance, often within a single question.

A different approach

For previous certifications, I relied on video courses and practice test platforms. They work, but they come with limitations: courses are linear for the most part and time-consuming, and practice tests often recycle the same question patterns without adapting to your weak areas.
But we live in a much different and exciting era now. From 2007 to 2018, advancements in cloud computing made computing power and AI infrastructure more accessible, leading to increasing adoption, innovation, and advancement in machine learning. In 2022, chatbots that use artificial intelligence (AI) and natural language processing (NLP) to have human-like conversations, like OpenAI’s ChatGPT, became widely known for their conversational abilities, renewing AI interest and development across the industry. Generative AI leverages deep learning and massive datasets to produce high-quality, human-like creative outputs. AI is revolutionizing industries, driving innovation, automating complex processes, and delivering exceptional user experiences at scale.
So why not take advantage of what we have at hand to assist in passing the exam, I thought.
So for this exam, I decided then to take a different approach. I used  Kiro  (agentic AI-powered Integrated Development Environment) with  Claude Sonnet 4.6  model (Anthropic’s frontier model with strong reasoning and instruction-following capabilities) as my primary study tool. I fed the model the  official SAP-C02 exam guide , domain breakdowns, and relevant AWS documentation, then structured my preparation into distinct rounds, each designed to build a specific skill needed for the exam.
I also used the  AWS Documentation MCP Server  and the  AWS Knowledge MCP Server  to query AWS documentation directly within my study sessions. These Model Context Protocol (MCP) servers allowed me to pull accurate, up-to-date AWS documentation into the model’s context without leaving my workflow. Additionally, I used the different  AWS service FAQ pages  as a supplementary source, as they often contain concise explanations of service behavior and limitations that are directly relevant to exam questions.
The reasoning was straightforward: the SA Professional exam does not test rote memorization. It tests your ability to parse complex scenarios, identify what matters, eliminate distractors, and select the architecturally sound option. These are skills that benefit from active practice with immediate feedback, which is exactly what a well-prompted AI model can provide.

My timeline

I originally scheduled the exam for 21 days out. Life happens, though. A customer engagement required me to travel for a week, which cut into my preparation time. I had to compress the plan into 16 days of actual study, which meant being more intentional with every session.
My AWS Solutions Architect Professional Exam Timeline (picture by author)

Before diving into structured practice, I reached out to my professional network. I talked to colleagues and peers who had already taken the exam or were deep in the services it covers. I asked them what surprised them, what they wish they had studied more, and what strategies worked for them on exam day. Their firsthand experience helped me shape my study plan and focus on the areas that mattered most. When preparing for an exam of this caliber, take all the help you can get.

Round 1: Keyword identification

Objective: Train the ability to rapidly extract relevant information from verbose scenarios
Because the exam questions are lengthy and can exhaust you mentally over three hours, I needed to develop the skill of quickly identifying the keywords and constraints that actually drive the correct answer. Most of the text in each question is context padding: realistic but irrelevant details designed to simulate real-world ambiguity.
Beyond the mental fatigue, this verbosity is designed to test your composure. At some point during the exam, the natural human response is to grow impatient, feel irritated by yet another wall of text, and rush through just to get it over with and move on to the next question. Candidates who give in to it start skimming, miss critical constraints, and select answers that sound right but violate a requirement buried in the middle of the paragraph.
This is exactly why I trained keyword identification as a deliberate skill. I prompted the model to generate exam-style questions. My task was to read each question and write down what I believed were the keywords. The model would then reveal the keywords in bold, allowing me to calibrate my instincts.
Example Questions: Keyword Extraction in Practice
#1 — “A company that has grown into one of the leading stock photo marketplaces in its region has built up a large collection of user-submitted stock photos contributed by a diverse community of amateur and professional photographers. As part of a broader effort to modernize its platform, an AWS Lambda function processes and extracts metadata from these photos to make a searchable catalog available to its customers. The metadata is extracted depending on several rules maintained by the data engineering team, and the output is sent to an Amazon ElastiCache for Redis cluster. The metadata extraction is done in several batches and the whole process takes about 45 minutes to complete. Whenever there is a change in the metadata extraction rules, the update process is triggered manually before the extraction process starts, which occasionally introduces delays when the on-call engineer is attending other meetings. The CTO has formally asked the Solutions Architect to implement a solution that addresses these concerns before the upcoming peak season. The engineering leadership has noted that any proposed solution should avoid replacing the existing AWS Lambda function or the Amazon ElastiCache for Redis cluster, as these are considered stable and well-tested components. As the stock photo submissions are steadily growing, the company wants to reduce the metadata extraction time for its catalog. Which combination of steps in the following options would meet this requirement most efficiently? (Select TWO)”
Notice how much of the original question is narrative filler: the company’s marketplace position, the diverse photographer community, the broader modernization effort, the on-call engineer’s meetings. None of that influences the answer.
The keywords reveal what actually matters: the current architecture (Lambda, ElastiCache for Redis), the constraint (keep these components), the problem (45-minute batch processing), and the goal (reduce extraction time).
Here is another example:
#2 — “A company operates an online auction platform using AWS Fargate containers behind an Application Load Balancer (ALB). The platform supports users across multiple regions and processes bids in real time. The company uses Amazon Route 53 to manage a public hosted zone for its domain. The domain routes traffic through an Amazon CloudFront distribution, which uses the ALB as its origin. Recently, users have reported seeing intermittent HTTP 502 (Bad Gateway) errors, primarily during peak evening hours. After initial investigation, the team discovered that a misconfiguration in the Fargate task definition is the root cause. The team estimates the fix will take approximately 45 minutes to deploy. As a result, users are presented with the ALB’s default error page. The company wants to improve the user experience by redirecting users to a custom error page while the team works on resolving the underlying issue. The solution must minimize operational overhead. Which solution would meet these requirements?”
Here, the question pads the scenario with details about the auction platform, Route 53 hosted zone management, and the investigation process — none of which influence the answer.
The keywords reveal what actually matters: the architecture chain (Fargate + ALB + CloudFront), the specific error (HTTP 502), the constraint (least operational overhead), and the goal (custom error page). Since CloudFront natively supports custom error responses for specific HTTP status codes, the solution leverages CloudFront’s built-in error page configuration — requiring no additional infrastructure or code changes.
📥Try Round 1 yourself: Download the exact prompt I used to train keyword identification with Kiro  here .

Round 2: Elimination technique

Objective: Develop systematic elimination of incorrect options
In this round, I asked the model to present questions with their answer options, and my task was to eliminate wrong answers before selecting the correct one. The model would then confirm which options could be eliminated and explain why.
Elimination patterns I trained on:
  • Service mismatch: An option suggests a service that does not address the stated requirement (e.g., adding a Security Group to solve a performance problem)
  • Constraint violation: An option replaces a component the question explicitly says must be retained
  • Over-engineering: An option introduces unnecessary complexity when a simpler solution exists
  • Scope mismatch: An option solves a different problem than what was asked
  • Self-contradicting options: An option that negates itself based on how the service actually works
This round was particularly valuable for this exam because the answer options are often architecturally plausible in isolation. The skill is recognizing which option most directly addresses the specific requirements and constraints stated in the question. To illustrate how this works in practice, here is how I applied elimination to the two questions shown earlier:   
Options for question #1 (Lambda and RDS connections during peak traffic): A. Introduce an Amazon ElastiCache for Redis cluster to cache data that is frequently retrieved from the RDS database. B. Optimize Lambda function memory settings and make sure database connections are properly closed once data has been fetched. C. Update the Lambda function to route database queries through an RDS Proxy endpoint. D. Fine-tune the Lambda function’s timeout configuration and raise the limit for concurrent executions. E. Activate Multi-AZ deployments on Amazon RDS to improve availability and enable automatic failover.
Eliminated (Scope mismatch):
  • B — While memory adjustments may enhance Lambda performance slightly, they do not directly tackle scalability or connection management issues during peak loads. Similarly, closing database connections after retrieval is a good practice but does not efficiently handle connection management in high-traffic scenarios.
  • D — Adjusting timeouts and concurrency limits does not address the core problem of too many database connections being opened simultaneously.
  • E — Multi-AZ improves availability and failover, not connection management or performance under load.
Selected: A and C
Options for question #2 (CloudFront HTTP 502 errors and custom error page) A. Upload an error HTML page to an Amazon S3 bucket that is configured with static website hosting. Modify the CloudFront distribution by adding a new origin for the S3 bucket and enforcing Origin Access Control (OAC). Create a CloudFront custom error response that returns the error page when the application returns a HTTP 502 error code as an origin failover. B. Create a Lambda function that adds a redirection rule to the path of the error page within the application. Configure a CloudWatch alarm to trigger this Lambda function when the Application Load Balancer (ALB) metric HTTPCode_Target_5XX_Count registers a value of one. C. Upload the error HTML page to an Amazon S3 bucket and enable static website hosting. Update the existing Route 53 record by adding a failover policy that routes traffic to the S3 website endpoint when health checks become unhealthy. D. Modify the Application Load Balancer’s rules by adding a condition for the path /*. Configure a fixed-response action that returns a HTTP 502 response code and a custom error message in the response body.
Eliminated:
  • B — Over-engineering. Introducing a Lambda function triggered by a CloudWatch alarm
  • C — Scope mismatch. Route 53 resolves to the CloudFront distribution. Failing over at the Route 53 level would bypass CloudFront entirely, breaking the CDN layer
  • D — Self-contradicting. Configuring the ALB to return a HTTP 502 with a custom message does not solve the problem
Selected: A
This round was particularly valuable for this exam because the answer options are often architecturally plausible in isolation. The skill is recognizing which option most directly addresses the specific requirements and constraints stated in the question.
📥Try Round 2 yourself: Download the elimination technique prompt with the full framework I used  here .

Round 3: Full Practice with Explanations

Objective: Simulate exam conditions and deepen understanding of incorrect answers
This may sound similar to traditional practice exams, but there is a key difference. With practice test platforms, you get a fixed pool of questions and a score at the end. With this approach, the model generated unique questions on demand, adapted difficulty based on my performance, and I could ask follow-up questions in the same conversation to go deeper into any topic. This was the most intensive round, and the feedback loop worked as follows:
  • If correct: a brief confirmation of why the answer is right
  • If incorrect: a detailed explanation of why my choice was wrong, why the correct answer is better, and links to the relevant AWS documentation
The documentation links were critical. When I got a question wrong, I would read the referenced AWS docs to build a deeper understanding of the service behavior, rather than just memorizing the correct answer.
Because Kiro maintains context across the conversation, the model tracked my accuracy across domains and identified patterns in my incorrect answers. It then generated additional questions focused specifically on those weak areas without me having to manually organize or request them.
📥Try Round 3 yourself: Download the adaptive practice prompt that generates unique questions and tracks your weak areas  here .

Round 4: Note consolidation

Objective: Create a concise reference from all study materials
In the final round, I provided the model with my various study documents, notes from Rounds 1–3, and asked it to generate consolidated notes organized by exam domain. This produced a focused reference document that I reviewed in the days leading up to the exam.
Key areas I ensured were covered in my notes:
  • AWS Organizations, SCPs, Control Tower, and multi-account patterns
  • Networking: VPC peering, Transit Gateway, Direct Connect, PrivateLink, VPN CloudHub
  • Storage decision trees: S3 classes, EBS vs EFS vs FSx, Storage Gateway types
  • Database selection: RDS vs Aurora vs DynamoDB vs Redshift vs Neptune
  • Migration services: DMS, MGN, DataSync, Snow Family, Transfer Family
  • Compute patterns: Lambda vs Fargate vs ECS vs EKS vs Batch
  • Analytics pipeline: Kinesis (Streams vs Firehose), Glue, Athena, EMR, Lake Formation
  • Security: IAM Identity Center, Cognito, KMS, Secrets Manager, WAF, Shield, GuardDuty
  • Cost optimization: Savings Plans, Reserved Instances, Spot strategies, S3 lifecycle policies
📥Try Round 4 yourself: Download the note consolidation prompt that generates exam-focused reference notes by domain  here .

Exam day strategy

Based on my preparation, I went into the exam with the following approach:
  1. First pass (0–120 minutes): Answer every question, flagging anything I was uncertain about. Target: 90 seconds per question on average for the first read.
  1. Second pass (120–170 minutes): Return to flagged questions with fresh eyes. Often, later questions provide context that helps with earlier ones.
  1. Final review (170–180 minutes): Quick scan for any unanswered questions or obvious errors.
Key tactics during the exam:
  • Read the question at the end first. In many cases, the actual question appears at the end of the scenario. Knowing what is being asked can help you filter the scenario more efficiently as you read through it.
  • Identify constraints immediately. Phrases like “without replacing,” “with minimal operational overhead,” or “most cost-effective” narrow your options significantly.
  • Eliminate before selecting. If you can confidently eliminate two options, your odds improve from 25% to 50% even if you are unsure between the remaining two.
  • Do not overthink. If a question seems straightforward, it likely is. The exam has a mix of difficulty levels.
  • Manage your energy. Take brief mental pauses every 20–25 questions. Three hours of intense reading is physically draining.

Key services to know deeply

Based on my exam experience, these services appeared frequently and require more than surface-level understanding:
Key services for the AWS Certified Solutions Architect - Professional (SAP-C02) Exam (picture by author)


What worked and what I would do differently

What worked

  • Structuring my preparation into distinct rounds gave me a portable study system — even during a week of customer travel, I maintained momentum because each round had a clear objective I could work toward from anywhere.
  • The keyword identification round was the highest-ROI activity. It directly translated to faster question processing during the exam.
  • AI-generated questions adapted to my weak areas meant I spent time where it mattered most, rather than reviewing topics I already understood.
  • Immediate feedback with documentation links created a tight learning loop that traditional practice tests cannot match.
  • Using Kiro with MCP servers meant I could query AWS documentation directly and control the model’s behavior precisely through prompts, ensuring questions matched the exam’s style and difficulty.

What I would adjust

  • I would start Round 3 earlier. The full practice questions with explanations provided the most learning per hour.
  • I would incorporate more multi-service architecture questions that span multiple domains, as these appeared frequently on the exam.
  • I would practice more with time constraints to better simulate the pressure of 2 minutes and 24 seconds per question.
One important recommendation: I strongly suggest taking the exam at a certification center instead of online. While online testing offers convenience, a certification center provides a controlled environment where technical issues and interruptions are less likely to happen during this important exam.
  • I would also recommend practicing in environments with some ambient noise — background conversations, white noise, or general activity. Testing centers are never completely silent, and even with ear plugs, you will still hear surrounding sounds. If you have only ever studied in perfect silence, this can disrupt your concentration on exam day. By training with a moderate level of background noise, you condition yourself to maintain focus regardless of the environment, so nothing catches you off guard when it matters most. In my case, my son playing and laughing outside my window was all the ambient noise I needed. :-)

Exam day & results

I arrived at the testing center in the morning, well-rested and focused. Three hours later, I walked out feeling cautiously optimistic but aware that this exam has a way of making you second-guess yourself.
I went back to work and tried not to think about it too much. AWS states that results can take up to five business days, so I was not expecting anything immediate. To my surprise, I received the email in less than 24 hours.
The moment I opened it and saw that I had passed, the relief and gratitude were overwhelming. All the prayers before every session, the early mornings, and the late-night study sessions over those 16 days had paid off.

Final thoughts

The AWS Certified Solutions Architect — Professional (SAP-C02) exam is a test of architectural judgment under pressure. It rewards candidates who can quickly parse complex scenarios, identify constraints, and select solutions that balance operational efficiency, cost, security, and performance.
Using generative AI as a study partner allowed me to prepare in 16 days with a focused, adaptive approach. The model did not replace the need to understand AWS services deeply, but it accelerated the process of building and testing that understanding through active practice rather than passive consumption.
Dedicating time to study while managing a career full of responsibilities, traveling to meet customers, raising a family, and handling life’s other obligations is far from easy. It requires real sacrifice — early mornings, late nights, and borrowed moments from an already full schedule. But the reality is, if we want to achieve meaningful professional growth, we need to commit the time. There are no shortcuts to genuine expertise, only smarter ways to use the time we carve out.
If you are preparing for this exam, consider how AI tools can complement your study approach. The key is structuring your prompts to build specific skills (keyword extraction, elimination, architectural reasoning) rather than simply asking for answers.
📥 Get all my prompts: I’ve open-sourced every prompt from my study plan. Clone the repo, customize the prompts for your learning style, and start practicing:  https://github.com/rnem/aws-sap-c02-ai-study-prompts.git 
The future of certification preparation is not about replacing human learning. It is about making that learning more targeted, adaptive, and efficient.
If you’ve already taken the exam, I’d love to hear from you — how long did it take you to prepare and pass? Let me know in the comments below! 👇🏻
Happy learning and good luck!
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article