Built by one engineer, in public

TensorUp is written by Ajay Kumar, a technical architect in agentic engineering: designing the systems that let AI agents do real work in production. He has spent thirteen years building production software, the last three with generative and agentic AI.

Most writing about AI describes what it can do. TensorUp explains how it works, what it takes to ship and how it breaks, across five topics: models and research, agents, AI engineering, tools and products, and teardowns of real systems. Every post has a diagram, and code when there is code to run.

How a post is made

Hands-on posts start as a working system in a public repo. The write-up comes after that system has run, and failed, often enough to say something true about it. Explainers and teardowns start from the primary sources: papers, documentation, release notes and the products themselves. Every claim is checked and every line of code is run before a post goes out.

TensorUp is written alongside a full-time role, so posts come when they are ready rather than on a fixed day.

Work with me

I take on a small number of writing and advisory engagements on AI architecture, agents and getting AI features into production. For sponsorship of the newsletter, a rate card will be published once the audience is large enough to be worth your money. Either way: hello@tensorup.ai.