About

I’m Ashish Pandey. I build machine learning systems for a living, and I write here about whatever I’m learning — mostly data, models and the machinery underneath them.

Why this site exists

Most ideas in machine learning aren’t hard. They’re badly explained.

They arrive wrapped in notation that assumes you already understand them, or flattened into a metaphor so loose it falls apart the moment you ask a follow-up question. Either way you come out the other side able to repeat the words, but not able to use them — and certainly not able to tell when the thing is about to fail on you.

So I write the explanation I wanted when I was learning it:

  1. The intuition first — one line you could repeat to a colleague over chai.
  2. Then an everyday analogy — concrete, with real numbers, and honest about where the analogy stops working.
  3. Then the actual mechanism — the maths and the code, once you already know what they’re for.
  4. Then where it breaks — the failure modes, the misconceptions, the things that bite you in production.

That last part is the whole point. Knowing where a system fails is worth far more than knowing where it works.

Everything here is written from things I’ve actually run. If there’s a number on this site, I measured it.

Say hello

I like hearing from people who are learning the same things — or who think I’ve got something wrong.