PatricioLumbe

Flagship project · AI + AWS · Serverless

AI Ops Serverless Platform

When something breaks in the cloud at 3am, this system does the noticing, the figuring out and the explaining automatically. Claude reads the alarm and writes a plain English diagnosis, then waits for a human to click "approve" before anything runs. Every possible action is pre approved in Terraform, never decided by the AI.

• 
• 

In plain terms

An engineer no longer wakes up to a wall of metrics and spends fifteen minutes working out what happened. The alarm arrives already explained, with a proposed fix waiting for a yes or no.

View project diagram

🇬🇧 Watch in English🇫🇷 Regarder en Français🇪🇸 Ver en Español🇵🇹 Assistir em Português

AI Ops Serverless Platform

Technical case study · PDF

View & download PDF

View the code on GitHub →

Google Cloud · Multi-cloud proof · Live right now

Cloud Portfolio Platform

The website you are looking at right now is itself a Google Cloud project. Firebase Hosting for global delivery, Cloud Storage for video assets, secrets kept in Secret Manager, an uptime check every five minutes, and a budget alert so cost never surprises anyone. Every piece is defined in Terraform.

• 
• 
• 
• 
• 

In plain terms

The whole site can be rebuilt from an empty Google Cloud project by running one command, and nobody has to keep a password anywhere for it to deploy.

View project diagram

Watch the walkthroughVisit patriciolumbe.com (live)View on GitHub

Demo video coming soon

Kubernetes & GitOps · Scales to zero · No kubectl in prod

EKS GitOps Platform

A Kubernetes platform where the only way to deploy is a Git commit, nobody types kubectl against production. A worker service scales down to zero pods when there is no work, so idle time costs nothing.

• 
• 
• 
• 
• 

In plain terms

You stop paying for servers that are doing nothing, and deploying is just merging code. Nobody needs console access to production to ship a change.

View project diagram

View on GitHub

Demo video coming soon

Infrastructure & Security · Multi-AZ · Zero static keys

Resilient AWS Infrastructure

An AWS environment designed to remain available during an Availability Zone failure, rolls back a bad deploy on its own before it becomes an incident, and runs CI/CD without a single stored AWS key anywhere.

• 
• 

In plain terms

If a data centre goes down or a release breaks something, the site stays up and the bad version is pulled back automatically, before it turns into an incident.

View project diagram

View on GitHub

More infrastructure, automation and lab projects on github.com/Lumbenlengo →

1. 01

   Private by default

   Application and database tiers live in private subnets across two availability zones. The only public entry point is the load balancer, sitting behind WAF rules. Nothing else is reachable from the internet, including the machines I administer it from.

   Demo video coming soon

2. 02

   No long lived AWS keys

   GitHub Actions assumes an IAM role through OIDC federation, scoped to this repository and branch. There is no AWS_SECRET_ACCESS_KEY to leak, rotate, or accidentally commit. That is the most common way small teams get breached.

   Demo video coming soon

3. 03

   Plan before apply, always

   Pull requests run terraform plan and post the diff for review. Only a merge to main runs apply, against remote state with locking so two pipelines can never fight over the same resources.

   Demo video coming soon

4. 04

   Recoverable, not just backed up

   Backups nobody has restored are a rumour, not a recovery plan. Because the whole environment is code, rebuilding it is a documented, repeatable procedure instead of an improvisation on the day of an incident.

   Demo video coming soon

Want the full write up with cost breakdown, trade offs and failure modes? Ask me for it →

1. Foundation · Early 2023

   AWS Cloud Practitioner certified

   I started where most people skip: understanding how AWS charges you and how it protects you, before touching automation. That is why I can look at an architecture today and tell you what it will cost to run and where it is exposed.

2. Architecture

   AWS Solutions Architect Associate certified

   The step from using AWS to designing on it. Multi AZ so one datacentre going down is not your problem, least privilege so a leaked key is not a disaster, and systems that stay up on a bad day instead of only on a good one.

3. Shipped · Open source

   Shipped a production ready AWS environment, publicly

   Empty AWS account to running application, fully in Terraform. Private by default networking, pipelines that authenticate without any long lived keys, and a plan reviewed before anything is applied. The whole thing is on GitHub, so you can read exactly how I work before you ever hire me.

4. Shipped · AI + Ops

   Built an incident response platform with AI in the loop

   A serverless system that notices an alarm, explains the root cause in plain English, and waits for a human to approve before it changes anything. Built on Lambda and Step Functions, every possible action locked down in Terraform. It runs for under a euro a month when nothing is happening.

5. Next

   AWS DevOps Professional

   Currently studying for it, alongside the next public build. The professional level covers advanced CI/CD automation, infrastructure orchestration at scale and operational excellence, which is the work I want to be doing next.