IT on Gal Fajon

CaraPaws: designing a UI prototype

contact@galfajon.dev (Gal Fajon) — Fri, 16 Oct 2026 12:00:00 +0100

I’ve been working with a teammate on an application named CaraPaws for the Global Classroom module at TU Dublin during my Erasmus exchange. It is a dog-care app built around a fairly simple idea: owners should be able to see how their dog is doing while somebody else is looking after them. With the UI design nearly finalised, I wanted to share how the prototype developed.

The welcome screen of the application.

Gradireland Graduate Careers Fair

contact@galfajon.dev (Gal Fajon) — Wed, 07 Oct 2026 12:00:00 +0100

I went to the gradireland Graduate Careers Fair at the RDS this week. It was a good chance to meet employers face to face, hear about different career paths, and come away with a few ideas for what to do next.

I attended two talks. Jill Barrett presented a session on LinkedIn and networking, which got me thinking about how I present my experience and make connections. I also went to the student volunteering talk. It was a useful reminder that what we do outside the classroom can help us develop skills and find out what interests us.

Open Source Science meetup on generative programming

contact@galfajon.dev (Gal Fajon) — Thu, 24 Sep 2026 22:04:28 +0200

Yesterday I went to my first open-source event in Dublin, an Open Source Science meetup on generative computing. The presentation centered around Mellea, an open-source Python library for building more structured LLM applications.

The main presentation focused on the idea of “generative programming”: treating LLM calls as parts of an otherwise regular program. Mellea gives you things like typed generative functions, requirements, validation and repair loops, and constrained outputs. I like the general direction a lot. LLMs are inherently nondeterministic, which is occasionally the entire point of using one, but that becomes a pain when you want to build software around them. Having a clearer boundary around that nondeterminism makes the resulting programs easier to understand, test, and reason about as a programmer.