An open course, built for the AI era

Build real modern data systems. Learn it your way.

Every modern data project runs from a single SQL query to a database spread across a thousand machines. You build it, scale it, and keep it correct and cheap, while AI writes the code.

Explore the course Start with SQL

Sign in with Google or email to save your progress. That is the only sign-up.

What this course teaches
Six module panels left to right, M1 through M6. The same database sits in every panel; each module adds the one capability the database needs at the next scale, from one machine to planet-wide.
A six-minute tour
Built for the LLM era

The model writes the query. The hard part is everything after.

A model writes SQL it can't vouch for. The query returns the wrong rows, or it runs fine on ten rows and falls over at ten billion, or it trusts data nobody checked. This course teaches you to catch that.

Direct the model, verify its work

Read, write, and debug SQL, and check the correctness and meaning of what an LLM writes for you. Agentic data stacks, and the checks that catch a wrong answer.

Real systems, torn apart

Case-study teardowns of UberEats, Claude and OpenAI agent memory, and OpenAI on Postgres. Real systems, not toy schemas.

Correct and cheap at scale

The same query can cost a cent or a thousand dollars. Storage layout, indexes, and query plans keep it right and fast at ten billion rows.

The throughline

Five questions every data project answers.

Strip away the titles and every data project answers the same five. They are design lenses, not modules, the same five from a notebook to a planet-scale service.

1. Correctness

Is the answer right, and can you prove it?

A right-looking query silently drops rows.

2. Efficiency

Fast and cheap at scale?

The same answer: a cent or a thousand dollars.

3. Live app

Data in, through, and out?

Thousands of ops a second, nothing lost or doubled.

4. Scale

Past one machine?

One database, many machines, still consistent and fast.

5. Agents

Build it with agents, trust it?

Build and run it with AI, then verify the result.

Flexible by design

How you learn it, and what it's for.

How you learn

Watch or read →

A 5-min video or a page.

Practice →

Problem sets and Colab.

Test yourself →

Quizzes, retake anytime.
What it's for

CS 145 exam →

Minimum viable prep.

Interviews →

Data, SQL, systems design.

Projects →

Build a real one.

Any way of learning serves any goal. Sign in to track progress.

Who it's for

For CS students, at Stanford and beyond.

Taking CS 145 at Stanford?

Start with how the class runs, then work through the modules in order. The weekly sections and problem sets keep you in step.

How the class works

Learning on your own?

Same course, no enrollment. Watch a video, take a quiz, or search a concept. Start anywhere.

Browse the course