Current interests & ideas

• AI & Machine Learning: LLM fine-tuning, data poisoning, and predictive modeling.
• Graph Theory & Network Science: Analyzing political networks (parliamentary voting behavior) and modeling epidemic spreads.
• Quantum Computing & Cryptography: Exploring the fundamentals of quantum information—such as running single-qubit circuits via the Felis framework and Alice & Bob’s cat qubit technology.
• Decision Systems & Strategy: Game theory applied to sports (tennis) or board games, and reinforcement learning.
• Reflections on AI & Sovereignty: Analyzing how Large Language Models shape public opinion and impact democratic stability. My focus is on the dominance of a few tech giants and how states use these tools to project influence and challenge national autonomy.

The hardware constraint

I run my experiments on a 2017 MacBook Air (1.5 GB VRAM, Dual-Core i5) without paid cloud services. This constraint encourages me to focus on small LLMs and lightweight techniques like LoRA. I prioritize deep understanding and efficiency over chasing massive scale.

Update: To fine-tune a 7B-parameter LLM with QLoRA, I used a T4 GPU (16 GB) on Google Colab

Closing Note

This is a space for curiosity and open-source collaboration. Everything shared here is open for you to reuse, adapt, or critique. If a project sparks your interest, feel free to dive in.

Photo by Pawel Czerwinski on Unsplash