The Newsletter
I've been working in AI, data, and engineering leadership for over a decade, currently as Director of Machine Learning and Data Engineering at the Wikimedia Foundation. Over the years, I developed the habit of writing down what I learn — not just from papers and blog posts, but from the messy reality of actually shipping AI systems. I called these my working notes.
These notes cover everything it takes to build and deploy AI in the real world. Because building AI systems isn't just about the model — it's about data pipelines, monitoring, cost optimization, team structures, and all the supporting infrastructure that nobody mentions in tutorials. It's navigating organizational dynamics, making build-vs-buy decisions, and understanding why things that work perfectly in development fall apart at scale.
This newsletter is my new working notes. Sometimes it's a question, sometimes a connection to something else I've read, sometimes something I realized halfway through a project or during a retro. These aren't glossy tutorials but artifacts of the thinking process — my record of shippable tactics, the real decisions and trade-offs that turn models into products. The patterns I've seen work, the anti-patterns I've seen fail.
If you’d like to follow along, subscribe.
The Author
I'm the Director of Machine Learning and Data Engineering at the Wikimedia Foundation, where I lead a team of engineers working on one of the world's largest websites. Before this, I was Director of Data Science at Devoted Health and held similar data leadership roles elsewhere.
I've written a hands-on ML book with O'Reilly, created bestselling AI flashcards that helped thousands learn ML fundamentals, published hundreds of tutorials, and hosted a popular data science podcast. I also founded a data visualization startup and built open source tools used by thousands. These days, I'm deploying AI at Wikipedia-scale while navigating the real-world challenges of costs, performance, and infrastructure.