Work
My capstone project at INESC TEC is getting deep fast. The focus is on using Graph Neural Networks to tackle the mapping problem — taking dataflow graphs and learning how to place them onto CGRAs. After getting a solid understanding of the CGRA architecture and the mapping problem, I've shifted to reading about Deep Q-Networks and how they're being used in state-of-the-art approaches. It's one of those rabbit holes that keeps getting wider the further you go. Good kind of overwhelming.
Building
The personal dashboard has come a long way. The big addition recently is a toolset for an LLM that can actually manage my app — scheduling, content, the works. It's less of a chatbot and more of an agent that has real control over the data. Still rough around the edges but the core loop works and it's genuinely useful day-to-day.
Next on the list: Deniz Cloud — a self-hosted home server running on a Raspberry Pi 5. The idea is a personal cloud with file storage, tiered SSD/HDD migration, PostgreSQL, MongoDB, an admin panel, and Cloudflare Tunnels for access. Already got port forwarding and DDNS sorted, now working through the foundation phase — monorepo setup, Docker Compose, database configs, and the auth system.
University
Things are moving along well. Every class has a project this semester, which keeps things hands-on.
Compilers — building a compiler for Java-- from scratch, which is genuinely cool. There's something satisfying about understanding what's actually happening between your source code and the machine. Ties directly into the GNN work at INESC TEC in ways I didn't expect.
Parallel and Distributed Systems — probably the course I'm enjoying most right now. Concurrency models, distributed architectures, all of it. The kind of content that makes you rethink every system you've ever built.
Graphical Computation and AI are rounding out the semester — less exciting so far but early days.