SUSHANT GUNDLA · LEARN← the site
← All courses

A hands-on course · 8 lessons

Build with LLMs

The working engineer's start.

You can already code. This is the other half — what the model actually is, how to make it do what you asked, how to give it your data, how to let it act, and how to prove it works.

Lessons run about five to eight minutes each and end with a short recall quiz, so you find out what stuck before you move on.

Part 1The model

01

What the model actually is

A next-token guesser with a fixed-size scratchpad. Everything else follows from that.

~7 min
02

Prompts that get followed

Instructions the model actually obeys — and why yours get ignored.

~7 min
03

Structured output

Stop parsing prose. Make the shape a contract the model has to fill.

~8 min

Part 2Giving it knowledge

04

Embeddings and search

Turning meaning into numbers, so the model can find the right paragraph before it answers.

~7 min
05

RAG without the hype

Wire your own documents into an answer, and know which half to fix when it is wrong.

~7 min

Part 3Letting it act

06

Tool calling

Let the model use your functions, and understand exactly who runs what.

~7 min
07

The agent loop

Build the while-loop that turns tool calling into an agent, and the budgets that stop it.

~8 min

Part 4Trust

08

Evals: proving it works

Replace vibes with a number that tells you whether your last prompt edit helped.

~8 min

Work through it in order the first time — each lesson leans on the one before it. After that, come back to any of them on their own.