It’s been a lot of fun building a tech/AI projects area at Unieuro with the technology students. Plenty of mistakes and wins so far, and I think they’re worth sharing. I’m a big fan of “learning by doing” and of keeping what I learn as open as possible, so here we go.

  1. I underestimated — by a lot — how much computer science students can deliver with a coding agent account. It’s something that stuck with me last year, seeing the quality of the projects delivered in my capstone course. It was Matheus Rosendo Pedreira who caught my attention, and ever since I’ve been thinking about this projects area. And here we are, a year later.

  2. I underestimated — by a looooooot — how much time I’d need to coordinate the projects and how much I’d still have to get my hands dirty. What I promised the projects was that I would deliver, whether or not the scholarship students could. I missed deadlines, I got the project management wrong, and many times I left the students “kind of on their own”. I want to train “forward deployed engineers” — whatever that is — and I ended up being absent. What I mean by that: being able to clearly define with the client what needs to be delivered and what the expectations are; making the delivery; and following the thing as it runs. A “duck”, as we say in Brazil — a jack of all trades who swims, walks and flies, but none of them very well? Yes, but I think it’s the best way to prepare for the future.

  3. Projects funded by scholarships come with their own complexity. The biggest lesson is that they “need” to be paid per deliverable. That’s how the bureaucracy understands it, and it’s not worth fighting. Switching to deliverables instead of time — a habit of mine from the agile world — has been, let’s say, painful. Matching the company’s expectations with the students’ anxiety and needs… is fun. This is definitely the one I’m struggling with the most.

  4. In the world of AI/agents, it’s very hard to have a well-structured project. It’s still a lot of trial and error, even more than a regular software project. Combined with point 3, it’s quite hard to plan far ahead. It’s been nice to think more in terms of hypotheses to test and how we’ll implement them, and to put that into the reports. If anyone has experience with this, I’d love to exchange ideas.

Soon we should open up for more projects. I want to thank Larissa Zandonade and Pedro Fogaça for joining this adventure! I hope to be back here soon with more news about the projects. :)

Originally written in Portuguese and published on LinkedIn on October 20, 2026. I translated it into English with the help of AI (Claude): it let me bring my LinkedIn posts to people who don’t read Portuguese without rewriting each one from scratch. I reviewed the translation to keep my original meaning and tone.