About

I am an AI guy. In my professional caeer, I build agentic AI platforms, orchestrations and frameworks.

AI has always been my passion, since about a decade ago now, when I first looked at the math behind machine learning and deep learning models. My thoughts on the subject are inspired by writings and works of Andrew Ng, François Chollet, Aurélien Géron, Andrej Karpathy, Sebastian Raschka, and many, many others.

My thinking in AI and system design has been influenced by my background in Physics, then Astroparticle Physics. What are the underlying symmerties of a system? What stays the same, when the same idea is interpreted across many different software implementations, agentic AI frameworks? What IS an AI agent, really? And is there a universal representation of an agentic system, that arises naturally from the shape of a requirement?

Then comes the matter of pushing AI systems to production. Which LLM to choose, and why? How to separate data and hyperscaler infrastructure from agent logic? Does the infrastructure influence agentic design choices?

In this website, I explore these topics.

But in my mind, AI is not just confined to commercial apps, or large scale enterprise workloads. Models that learned from words can now look at a photograph, read a handwritten page, and help search the sky. They can see every sign on a seal from Harappa, and still no one can read it. I write about how they see: how they interpret and interact with the world, and where it stops.

This website is about AI, but in a way, it is about living with AI too. For the past years, I have worked extensively with modern LLMs and VLMs, across all model providers. I work every day with models from most of the major labs, open and closed: today Gemini, Claude, GPT, Qwen, GLM, DeepSeek, MiniMax and MiMo. Tomorrow, they will be something else. The research, the reading of source code, and the writing here are done together with them.

I think AI has made thinking easier: not because it thinks for us, but because we no longer have to think alone. I also write about how these models behave in real work, the things a benchmark score doesn’t show.

The models read, search, and gather what they can. The synthesis, the insights, and the learnings are mine.

I hope you enjoy reading about AI as much as I do writing about it.