About — Alexei Ferreira

Founder, Lx8 Labs. This page is the long-form personal bio. The studio overview lives on Team; the company front lives at lx8labs.com.

In one paragraph

I've been a DevOps engineer for 13 years. Started in 2013 in Brazil, moved to Berlin, ran production for fintechs and platforms, built K8s clusters for teams of 5 and teams of 500. About three years ago I started moving into AI engineering — not the demo kind, the boring infrastructure underneath that keeps an AI feature from embarrassing you when it ships. Lx8 Labs is the studio I'm putting all of that into: consulting work, the courses I always wished existed, and a couple of products I actually use myself (Tupã IDE among them). On the side I write physics papers — there's a Theory of Everything project under peer review at Foundations of Physics as I type this.

Path

2013

Started DevOps consulting in Brazil

First Linux SRE jobs in São Paulo. Founded Lx8 as a personal vehicle for freelance work that year.

2016–2020

Platform engineering · fintech, e-commerce

K8s, AWS, Terraform shops. Built and operated production. Got good at incident response, observability, and saying no to half-baked architectures.

2020

Moved to Berlin

Different climate, same containers. Started thinking seriously about what a small independent studio could look like.

2023

Pivoted toward AI engineering

LLMs hit a quality threshold where they were worth productionizing seriously. Started shipping AI features and learning what the boring layer underneath needs to look like.

2024

Tupã IDE prototype

Frustrated with Electron IDEs. Started a native macOS editor in Swift + Rust + Metal. Still in active development.

2026

Lx8 Labs as a studio

Consolidated consulting + courses + products + research under one roof at lx8labs.com. Submitted the bipartite-theory paper to Foundations of Physics in May.

What I do well

Taking systems people are afraid of and making them legible. The fintech that couldn't release on Fridays, the K8s cluster that worked but no one understood, the AI feature that demoed great and broke under 10 concurrent users. The hands-on, no-bullshit middle layer between "what does this metric mean" and "how do we ship this safely tomorrow."

Stack I'm fluent in

Kubernetes Terraform AWS GCP Python Go TypeScript Rust Swift PostgreSQL GitHub Actions Helm Argo Prometheus · Grafana OpenTelemetry LLMs · vLLM, TGI, Anthropic, OpenAI Three.js

What I write about

Lately: AI cost control, prompt versioning, the gap between "model accuracy" and "user-visible reliability." Older posts in the queue: Kubernetes failure modes nobody warns you about, deploy pipeline rewrites, the consulting business itself. Insights section on the homepage carries the queue.

Theoretical physics, briefly

Independent research on a bipartite reading of the Einstein–Maxwell system — organizing light–gravity coupling into three orthogonal channels (gravitoelectric, gravitomagnetic, Λ-cosmological). A paper went to Foundations of Physics (Springer) in May 2026. Whether it lands or not, the work lives in the open. The interactive 3D version is the "Light, Gravity & Time" visualization on this site.

Outside engineering

Hiking, Latin guitar (bad but enthusiastic), Brazilian Portuguese as a first language, German learned mostly from neighbours' kids. Believe small companies built on long relationships beat fast-growth ones run on investor patience.

Want to work together?

Open for consulting engagements (DevOps audits, AI feature productionization, platform reviews) and private cohorts of any course on this site. I work with teams of 1 to ~50 engineers and prefer fixed-scope projects to retainers.

Send a brief → Course catalog GitHub LinkedIn