My experience taking the AWS Certified Generative AI Developer - Professional

Agustín Villagrán — Fri, 03 Apr 2026 14:57:43 +0000

I recently passed the AWS Certified Generative AI Developer - Professional.
But what helped me the most wasn't just studying — it was building with Bedrock in production at the same time.

This isn't an official guide or a syllabus summary.
It's what I saw, what worked, what I'd change, and what's coming next.
Builder to builder.

Some context

Before this exam I already held 4 AWS certifications:

Cloud Practitioner (CLF)
AI Practitioner (AIF)
Developer Associate (DVA)
Solutions Architect Associate (SAA)

The Gen AI Dev Professional was my first Professional-level cert, and I planned it knowing the jump in difficulty is real.

But here's what I think made the biggest difference: while studying, I was simultaneously building GenAI projects with Amazon Bedrock at work. Not side projects or labs (which also help a lot) — real production systems and POCs with Bedrock, SageMaker, OpenSearch, DynamoDB, S3, ECS Fargate, Lambda, among the main services.

That gave me something no course can: real context and hands-on learning.
When the exam asked about RAG flows, model invocation, or guardrails, I wasn't just recalling theory — I was recalling architecture decisions I'd made in real projects.

What I saw on the exam

Before preparing, I read a post by Ana Barragan: "Notas de alguien que ya paso el AWS Certified Generative AI Developer - Professional". Her notes helped me calibrate my study, and I can confirm almost everything she said from my own experience:

RAG is the heart of the exam.
Ana estimates ~70% of questions relate to RAG. I agree. It's not always explicit, but the pattern repeats: context retrieval, embeddings, chunking, response quality. If you understand RAG well, you're already most of the way there.

Bedrock is the entry point to everything.
Knowledge Bases, Guardrails, model invocation, Agents — it all flows through Bedrock. Having used it in production gave me a huge advantage here.

Guardrails matter more than you'd think.
In several questions, Guardrails was the piece that made the solution viable in production. It's not a "nice to have" — it's part of the design.

Chunking: if the model doesn't understand the context, the problem is probably there.
Multiple questions where the fix wasn't changing the model or the service, but improving how information is split. Large vs small chunks, overlap, impact on relevance and latency. Ana's mental rule is very efficient.

SageMaker shows up, but far less than RAG.
Only when the scenario requires fine-tuning, custom datasets, or MLOps. If it doesn't say that, the answer is probably Bedrock.

Lambda + API Gateway are the glue.
They appear as part of the flow, not the focus. API -> Lambda -> Bedrock -> context retrieval.

CloudWatch: if it's in production, it needs to be there.
Observability always shows up in enterprise scenarios. Logs, metrics, alarms.

What I'd change if I studied again

1. More practice exams, much earlier.
I studied theory first and left practice exams for the end. Mistake. Practice exams aren't just for measuring readiness — they're the best learning tool. They show you the format, the language, and the reasoning patterns AWS expects. If I started over, I'd take practice exams from week two.

2. Classify errors after every practice exam.
It's not enough to see what you got wrong. You need to classify each error: unknown service? Misunderstood concept? Misread the question? Each type of error has a different fix.

3. Time management: the hardest lesson.
I ran out of time. I left 4 questions blank.
That shouldn't happen, and the fix is simple: two-pass strategy. First pass: answer everything you're confident about and flag the ones that gave you doubts. Second pass: return to the flagged ones. Never leave a question unanswered.

Also: request the ESL accommodation (+30 minutes). It's free, it's for non-native English speakers, and those 30 minutes can be the difference between passing and failing.

4. Skill Builder + Udemy, not one or the other.
Ana recommends AWS Skill Builder, and she's right — it uses the same language as the exam and presents very similar scenarios. I primarily used Stephane Maarek's Udemy courses for theory and practice exams. The ideal combo is both: Udemy to understand and practice, Skill Builder to think the way AWS expects.

Is it worth it for your career?

People ask me this often.

My short answer: it depends on where you're starting from.

If you're already working with AWS and GenAI services in production, this certification organizes and validates what you already know. It gives you a common vocabulary with your team and clients. And in many contexts, a Professional-level cert on your resume carries more weight than you'd expect.

If you're just starting out and haven't touched Bedrock or RAG yet, the certification alone won't change your career. But it does give you a clear map: what to learn, in what order, and which services matter in production.

What I can say for certain: the process of studying for this exam made me better at my job. Not because the exam is perfect, but because it forces you to understand why each service works, not just how to use it.

The certification doesn't replace experience. But if you pair it with real projects, they reinforce each other.

What's next

The Gen AI Dev Professional was my first Professional certification, but it won't be my last.

Next up is the AWS ML Engineer Associate, starting in April. After that comes the Solutions Architect Professional — the hardest AWS exam (45% pass rate).

The goal is to build a certification portfolio with depth, not just quantity. Each cert connects to the previous one: AI Practitioner -> Gen AI Dev Pro -> ML Engineer -> Solutions Architect Pro — reinforcing and expanding what I already know.

And most importantly: I'm still building in production with these services. Certifications validate what you know, but real experience is what makes you absorb the knowledge by putting it into practice.

Thanks

I want to close by thanking Ana Barragan for her post. Reading it before my exam helped me focus my study where it actually mattered, and several of her observations were confirmed point by point in my experience.

If you're preparing for this certification, I hope my notes add some clarity.
And if you're hesitating about taking the step: schedule the exam. Until you set a date, the studying never ends.

Looking for spanish version?

Thanks for reading!
Great day!