I architect software systems end-to-end: from service boundaries, APIs, and data flows to cloud infrastructure, delivery pipelines, and day‑2 operations. AWS is the platform I use most, but my core value is software architecture under real constraints: reliability, cost, security, latency, maintainability, and teams that need to ship. Lately that is mostly applied to Generative AI applications.

I’m a Sr Innovation Architect (Office of the CTO) at Caylent. I also write Simple AWS, a weekly newsletter read by 45k+ engineers, architects and team leads who want clear explanations and practical patterns, not copy‑pasted architectures.

Advisory & consulting (where I’m most useful)
- Architecture reviews and “what should we build next?” roadmaps (software + cloud, GenAI)
- High-stakes tradeoffs: build vs buy, serverless vs containers, async vs sync, data modeling and boundaries
- GenAI solution architecture: production patterns, evaluations, model selection, production deployments and best practices, cost/latency tradeoffs, and integration into existing products
- Cloud-native architectures on AWS (serverless, event-driven, containers where it fits, networking and security)
- Delivery practices for code (Python, JavaScript/TypeScript) and Infrastructure as Code (CDK, Terraform, CloudFormation)
- Enablement: workshops, presales support, and technical talks for teams and leadership

Credibility (high level)
- 11+ years building software professionally; 25+ projects across startups, enterprises and SMBs
- AWS Solutions Architect Professional; AWS Community Builder and Builder of the Year award winner; former AWS Authorized Instructor
- International speaker (Argentina, US, Mexico, Colombia, and looking to expand this list)

I’m not actively job-hunting, but I’m open to selective advisory/consulting engagements (reviews, workshops, short-term roadmap work, or fractional architecture and technology support). If you think I can help, DM me.

Start reading here: simpleaws.dev
Sponsorships for Simple AWS: https://www.passionfroot.me/guilleojeda

This is very important. In fact I'm preparing my own version of this talk. Plz go read it

undefined undefined

Here's the link to sign up: https://learn.simpleaws.dev/offers/fNm3CMAr

Looking forward to meeting you there!

Simple AWS Newsletter

Join over 3000 devs, tech leads, and experts learning real-world AWS solutions every week with the free Simple AWS newsletter.

Sign up for free at https://www.simpleaws.dev

AWS Made Simple opening chapter Microservices

For a limited time, I'm giving away my book AWS Made Simple and Fun.
The opening chapter is Microservices Design, and it gets even better after that.
Get it here: https://www.simpleaws.dev/aws-made-simple

Want to learn AWS Simple AWS

If you want to learn AWS, join my free newsletter, Simple AWS.
You'll learn a new solution every Monday, going from the basics to all the best practices.

Subscribe at https://newsletter.simpleaws.dev

Simple AWS Here's the link

Here's the link: https://simpleaws.dev
You'll get a free issue every Friday.

See ya there!

AWS Made Simple Subscribe

Subscribe here: https://www.simpleaws.dev/aws-made-simple
And you'll get the download link in your inbox!

AWS services

I'm available for consulting if you need help with your cloud architecture, or guidance with an issue on AWS.

DM me, or find out more at https://www.guilleojeda.com

Haikus Simple AWS

Do you like AWS and haikus? I run a free newsletter called Simple AWS, where you get a haiku ever Monday and Thursday. You also get a tl;dr about a service, some tips and a recommended tool.

Interested? Sign up at www.guilleojeda.com

See ya there!

Simple AWS link

All links

By the way, here 
Do you like AWS and haikus? I run a free newsletter called Simple AWS, where you get a haiku ever Monday and Thursday. You also get a tl;dr about a service, some tips and a recommended tool.

Interested? Sign up at www.guilleojeda.com

See ya there!

Simple AWS ECS

Here's the link: https://newsletter.simpleaws.dev/p/simple-aws-4-ecs

Simple AWS helps you build better solutions on AWS. Runs for free every Monday, featuring a use case, a solution, advanced strategies and a list of resources.

Guille Ojeda services

If you need help simplifying solutions, here's what I can do for you: www.guilleojeda.com/services

Let's talk, no obligation.

Simple AWS Community

You can join the Simple AWS Community at https://learn.simpleaws.dev/offers/fNm3CMAr

Engage with experts, grow your knowledge, and take your career to new heights. Sign up today and get the first month free!

dreamoncloud

de-cero-a-recruiter

moar-clients

moarclients

Most generative AI projects don’t fail in the demo; they fail in production.

The gap between a great demo and a successful deployment often comes down to what teams choose to prioritize. At Caylent, we’ve deployed over 200 generative AI applications across real-world enterprise environments, and along the way, we’ve learned what actually separates successful systems from costly experiments. In our new blog by Guille Ojeda, we break down the 6 lessons that matter most, from defining clear inputs and outputs to building strong evaluation frameworks, designing for real user constraints, optimizing UX, and keeping costs under control.

Read it here 👉 https://hubs.li/Q03_Qvk_0

Sr Innovation Architect | AWS & GenAI | Creator of Simple AWS (45k+ readers) | Speaker

🌶️ take: code quality matters less now. Don't write code a human can maintain in 10 years. In 2-3 years we'll use AI to improve it.

AI can already improve 💩 code. We're constantly using it to modernize legacy applications. And that's with 2025 models and tools.

Consider where we were 3 years ago. CodeWhisperer 🤣, Copilot just being an autocomplete, GPT-3 hallucinating everything. Now consider how far we've come in 3 years, and think about how far we'll come in 3 more years.

No, we won't reach AGI.

And maybe we still won't reach a point where AI writes better code than you.

But we will reach a point where AI can maintain and improve 💩 codebases easily and consistently. 

I'm not claiming we should generate 💩 code now, to be clear. What I'm saying is that the maintainability horizon for code is no longer 10 or so years, like we're used to, like we've been taught. It's 3 years at most.

And even if in 3 years we're still coding, coding will look nothing like it did 10 years ago, when we learned these metrics.

So stop obsessing over code quality. Pay attention to it, for sure! But redo the math about the cost of maintaining software. That cost is going down hard.

My tokens are back!! I can finally code again!
And I'm already down to 91% 😨
Pro tip I just discovered: The 7 days start counting from when you use your first token, not from when you get them back. I got them back around noon and sent my first prompt around 3 pm, now it says 3:11 pm.

I can't believe I'm timing the refresh on the usage limit.

Guille Ojeda

Talks about #aws, #cloud, #startups, #consultants, and #softwareengineering

Sr Innovation Architect | AWS & GenAI Architect | Speaker | Creator of Simple AWS (45k+ readers)

Code coverage is a lie.
You can hit 90%+ line/statement/branch coverage and still have tests that pass in the presence of bugs.
Mutation testing is one of the most practical ways to expose that.

Mutation testing intentionally introduces tiny "bugs" (mutations) into your code, then runs your test suite against each mutated version.

If your tests fail, it means they succesfully detected the bug. For some odd reason, in mutation testing we say they killed the mutant. If your tests pass, they failed to detect the bug, and we say the mutant survived. You end up with a mutation score = mutants killed / total mutants, that measures how effective your tests are.

Mutation tools apply common fault patterns across your code, like:
🐞 > becomes >=
🐞 == becomes !=
🐞 && becomes ||
🐞 if (x) becomes if (!x)
🐞 return values replaced (e.g., returning 0, null, false)
🐞 method calls removed / skipped
🐞 arithmetic operators swapped (+ to -, etc.)

These are the kind of mistakes real humans make. Say you have:

int applyDiscount(int price, boolean vip) {
  if (vip) return price - 10;
  return price;
}

A mutant might flip the condition: if (!vip) return price - 10;

If your tests only check that "VIP gets a discount" but never check that "non‑VIP does not get a discount", this mutant will survive.

Mutations are performed automatically by tools like Stryker (JS/TS) or mutmut (Python). You pick a module, run some baseline tests to make sure your test suite works, then you generate mutants and execute the tests. You'll get a report, and then you have to investigate the surviving mutants to figure out why they survived.

There are 3 main reasons why mutants survive:
- You're not testing for what was changed in the mutation --> Add tests
- Your tests are coupled to an implementation detail instead of to behavior --> Always test behaviors, not implementation details
- The change from the mutation doesn't alter any tested behavior --> This may be dead code or untested code, but it's also possible it's just a syntactic change which doesn't alter semantics. Tools let you suppress these surviving mutants. This is normal.

Code coverage is about how many parts (lines, statements, branches) of your code have tests. Mutation testing reveals how many potential changes to your code are not covered by tests. If you're writing tests as a way to detect regressions and loss of quality as you make changes to your code, the second is more important.

Post inspired in one by Paul Hammond, who knows a lot more than me about mutation testing.
And the image is from Marvel's House of M.

From our upcoming 2026 GenAI Outlook whitepaper:
"The most critical risks introduced by agents are not really “bad reasoning” or hallucinations, no matter how much people on LinkedIn complain about that. The new critical risks are about good reasoning + too much authority."

Not sure the shoutout to all of you LinkedIn complainers will make it to the final version 🙃.

Basically, it's time to stop complaining about LLMs hallucinating or getting it wrong. We get it, it happens. And it's going to continue happening, no matter how many LinkedIn posts we make about it. Either we stop using LLMs, or we set good guardrails.

And what's changing in 2026 is that we're giving agents a lot more power over stuff. Guardrails for that look different. It's not just about reviewing the command an LLM outputs, it should be mainly about the execution of that command, and the scope and blast radius of the actions the LLM can take. And about the permissions we give it.

If it takes action, it's a principal. If it's a principal, least privilege.

This is from the 2026 Generative AI Outlook whitepaper by Caylent. We're publishing it second half of January. In the meantime, I'll keep leaking stuff here 😁

Alright, I guess I can't work until Thursday 🤷

Just kidding, I have a lot of writing and budgeting to do 😭.
The good news is that ChatGPT still works even if you hit your limit in Codex.
The bad news is that I'm on the cheap Claude subscription, there's no way it'll keep up. And I'm too cheap to pay for the expensive one

taplio

Guille Ojeda's Linkedin Analytics

Guille Ojeda

Performance ScoreGuille Ojeda's LinkedIn Performance

What is Guille talking about?

Who is engaging with Guille

Guille Ojeda's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Guille Ojeda's Linkedin Analytics

Guille Ojeda

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Guille Ojeda's LinkedIn Performance

What is Guille talking about?

Who is engaging with Guille

Guille Ojeda's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Guille Ojeda's Linkedin Analytics

Performance ScoreGuille Ojeda's LinkedIn Performance

Guille Ojeda's Best Posts (last 30 days)