$cat ~/posts/joining-sita

bilingual · en / pt

← /blog

joining SITA

2026-09-09·en·4 min read·careerai-engineering

Some news I am genuinely happy about: on June 3rd I joined Sidia, and I am now past three months there. I am part of the SITA team.

There is a specific kind of happiness when a team's scope reads like your own history: the feeling that you have been rehearsing for it the whole time without knowing.

the scope is the thread

SITA splits into a handful of teams: an AI team, a web team, a desktop team working in C#, a QA team, and a mobile team building for Android, plus, inevitably, people whose skills cross all of it. I am on the AI side. The common denominator across the team: building systems that automate what people used to do by hand.

Reading that split felt strange in the best way, because my own career is scattered across exactly this kind of ecosystem, even if I never worked the C# or Android side of it directly. I have analyzed data until it confessed something, poked at networks, worked the offensive side of security, and always, at the end, written the software that automates the whole thing. Landing on the AI team of a place shaped like this feels like the thread finally being pulled taut.

I will not talk about what we build, that stays inside. But the shape of the work I can say out loud: it is exactly the shape I like.

from js/ts to python + AI

For most of my career my stack was JS/TS and web in general, but the AI side was never really a separate thread, it was usually wired straight into the web work: A-Garra, a project of mine, and before that the data treatment pipelines I built during a university extension project. Without planning it, I had been building a portfolio with a near-perfect fit for a team that needed exactly that combination.

Python was never the new part of this: it was already my primary language in offensive security and data analysis, long before web became my day-to-day. What changes now is that this combination became the actual job, not a curious detour between other deliverables.

research as a direction

The bigger realization: research and development is the road I have decided to stay on from here forward. Doing science for real, not as a metaphor. The AI scope is where that is going to happen: ML, DL, NLP, computer vision, LLMs, the whole family.

It is already concrete: my TCC (the final undergrad thesis we do in Brazil) is pointed at this space, and a master's and a doctorate in projects around these themes come after.

the plan

Since August 1st, 2026 I have been running a study track I decided to call "the Plan", capital P on purpose. It started from a simple discomfort: my Information Systems degree never gave me the math foundation AI engineering actually needs, and since I cannot go back in time, I decided to chase that foundation on the side. It is a five-stage track, Coursera-only (the access came through Sidia and expires in June 2027), with one goal: actually becoming an AI/ML Engineer. The first credential is already out, the IBM Generative AI for Software Developers Specialization, approved on August 23rd.

There is a lot more decision-making, cutting and math behind this track than fits here. It deserves a post of its own, practically a manifesto, and it is already drafted.

forged in fire

That road is already bumpier than it sounded on paper. I picked NLP as the first serious stop on it, mostly because spaCy was already part of the day job. It went fine until it very much did not: I could get to the right output without understanding why, and the reason turned out to be that I never got a solid foundation in data structures and algorithms in college. A spaCy Doc is basically a DSA problem wearing an NLP costume, and I only found that out by hitting the wall face first. What I am doing about it is a story for the next post.

A small coincidence I like too much not to mention: I started my career as a dev on November 25th, 2025, and I am now past the nine-month mark of that. Over nine months as a dev, a bit past three months at SITA, the thread keeps pulling.

The first post on this blog says writing in public forces clarity. Consider this the first commitment published under that rule.