About

gojune21 is a blog documenting the real problems encountered while designing AI and data systems. It covers the pitfalls and design principles that commonly surface when building data pipelines, LLMs, and AI agents. The focus is not on product- or vendor-specific usage, but on universal architecture principles that hold across any tech stack.

Who writes this

This blog is run by a practitioner who designs and builds enterprise AI projects at a large organization. The starting point for each post is a problem hit while building AI pipelines on real organizational data in production. It shares none of the specifics of any company, client, or internal tool — only the universal design lessons drawn from them.

The perspective

The posts here are not stitched-together theory; they start from hands-on experience building systems. The moment a summary swallows a weak but important signal, situations where you must design operations on the premise that probabilistic merging is never 100% correct, ways to use AI without sending sensitive source text outside your boundary — these are the points where I actually got stuck and worked my way out. As a result, the writing tends to be narrow and deep: the details larger outlets skip, written down together with why those details actually mattered.

What readers get

The primary readers are enterprise architects and developers. If you've hit the same wall while designing AI and data systems, you can take away the name for that problem and the design principle for dealing with it. Each post aims to answer one concrete question self-containedly — so that even if you arrive from search and read just one, you leave with something.

Contact: [email protected]