About Mitch
I'm an independent data and AI engineer. I have a PhD in Computer Science, and since 2017 I've been building the systems businesses run on: behavioural tracking, data warehouses and pipelines, machine learning, marketing attribution, and more recently private AI, LLMs and retrieval-augmented generation.
I'm a hands-on operator, not a theoretical strategist. I design the architecture, write the code, and keep the systems running in production, on Linux, Kubernetes and cloud infrastructure. I've done this inside established B2B and B2C businesses, including high-volume ecommerce and advertising environments, so I've seen what survives contact with real traffic and real budgets.
The thread running through the work: businesses should have control over their data, infrastructure and AI, rather than becoming dependent on an expanding collection of proprietary platforms and vendors. Not every tool needs replacing and not everything should be self-hosted. But the decision about what you own and what you rent should be yours, made with full knowledge of the trade-offs.
I deliver consultancy and engineering work through Pragmatik, the business I founded. This site is me: the writing, the thinking, and the work.
How I think about technology
Ownership
Your data is the one asset a competitor cannot buy, and it deserves the same protection as anything else the business would fight to keep, not sent to whichever AI company has the slickest chatbot this quarter. That is not an argument against cloud services or SaaS. It is an argument for deciding deliberately what you own, what you rent, and what leaves your hands.
Practical engineering
Technology choices follow the problem. The smallest suitable solution beats the fashionable one, and a boring tool that runs for years beats a clever one that needs a babysitter.
Commercial value
A system earns its keep by improving how the business operates, decides, or makes money. Infrastructure for its own sake is a hobby, and clients should not fund hobbies.
The tools I reach for
The stack varies by problem, but this is the kit that keeps earning its place in production:
Want the longer version?
The journal documents the builds, the decisions and the costs. Or skip ahead and talk to me about your business.