BSc Computer Science & Engineering student building across full-stack web, edge AI, and Qt/C++ systems

I care about owning the code — what runs, why, and where the bytes go. While software is often generated fast and understood slowly, I'm betting on real engineering understanding, deliberate design, and software that holds up under pressure.

🌐 thedaxen.com

I'm developing with Qt, QML, and C++ toward embedded, automotive, and medical systems, where performance, constraints, and control matter. I want to understand the memory and execution model rather than stop at the highest abstraction.

My current Qt project is WardView, a responsive interface concept for a bedside patient monitor.

I'm going deeper into AI engineering: integrating and evaluating models, designing reliable inference workflows, and taking them to the edge, where latency, privacy, and resource constraints matter. I'm also leaning into Python and automation — wiring pipelines and building tools that handle repetitive work reliably.

My current learning project is LLM Internals Lab, where I'm implementing the internals of a GPT-style language model from scratch with Python and PyTorch. It covers tokenization, next-token data sampling, token and positional embeddings, and trainable self-attention, with unit tests validating each stage and its tensor shapes.

I've built real things with Next.js, React, and TypeScript — enough to have felt where shortcuts bite. I'm building engineering discipline while it's a habit rather than a rescue, and shipping polished, maintainable applications.

I use AI to move faster — never to replace understanding or ownership. I avoid vibe coding at all costs. The vibes can attend the planning meeting; they just don't approve the architecture.

I ground AI-assisted work in the codebase with specifications, ADRs, typed contracts, reviewable changes, automated and manual testing, and human review. I also take on no-AI challenges to strengthen the fundamentals.

I'm starting to contribute regularly to open-source projects through focused fixes I can understand, test, and own properly. I plan to keep contributing and gradually take on larger issues.

Upstream project	Community	My merged contribution
IPython		Restored pathlib.Path support in IFrame while preserving HTML escaping, with regression coverage.

Web         Next.js · React · TypeScript · modern fullstack
AI/Edge     Python · product integration · evaluations · model deployment · edge inference
Automation  Python · scripting · pipelines · model-in-the-loop tooling
Systems     C++ · Qt · QML · embedded · memory-aware programming

I'm focused on the Finnish and wider Nordic tech scene. If you're working on engineered web applications, AI-native products, edge computing, or Qt/C++ systems, I'd be glad to talk.

• Website: thedaxen.com
• Email: farzam.sadegh@gmail.com