System Design Interview Crash Course

It is a 2-week sprint through 30 complete system design problems. Each one is a full end-to-end design of a real system, from requirements and estimates through the API, data model, core flows, caching, storage, scaling, reliability, security, and the bottlenecks an interviewer will push on. The problems are grouped into 14 themed days, 2 or 3 a day, so you build breadth quickly without losing depth on any single design.

Engineers with a system design interview coming up who want volume and range rather than another explanation of what a load balancer does. It fits mid-level and senior candidates preparing for FAANG and other top-tech loops, engineers returning to interviewing after a few years in one codebase, and anyone who has read the theory and now needs to practise applying it to real systems under time pressure.

Grokking the System Design Interview is the foundation course: it teaches the concepts and the framework in depth, with video walkthroughs and 16 case studies. The Crash Course is problem-first, with 30 designs back to back on a 14-day schedule, written to be worked through rather than watched. If system design is new to you, start with Grokking the System Design Interview or System Design Fundamentals. If you already have the fundamentals and an interview in two weeks, start here. Plenty of learners use both, one to learn the method and one to drill it.

It is the fixed order that every design in the course follows: scope the problem, functional requirements, non-functional requirements, back-of-the-envelope estimates, API design, high-level architecture, data model, core flows, caching, storage, scaling, reliability, security, and bottlenecks. A dedicated lesson in the introduction explains what each step is for, why the order matters, and which steps to compress when you have only 45 minutes on the clock. Because all 30 problems run in the same order, the sequence becomes automatic long before you finish.

YouTube, Netflix, WhatsApp, Discord, the Twitter timeline, Uber, Airbnb, Amazon S3, the Amazon shopping cart, Stripe, a stock exchange, Google Search, Google Ads, an ad click aggregator, ChatGPT, Google Docs, Miro, Twitch chat, typeahead and autocomplete, Facebook People You May Know, LinkedIn connections, a metrics and monitoring system, a code deployment system, AWS Lambda, an API gateway, a distributed cache like Redis, a key-value store like DynamoDB, a distributed rate limiter, a distributed lock manager like Chubby, and a distributed job scheduler.

The built-in study plan runs 14 days at about 2 hours a day, which carries you through all 30 problems once. That first pass builds recognition. The course then walks you through a second pass, where you redraw each architecture from memory and read the deeper sections on caching, scaling, reliability and security that you skimmed the first time. Counting both passes and those deep dives, most people spend a good deal longer than the 14-day plan alone. It is fully self-paced, so you can compress it into a week of long sessions or spread it across a month.

You need to be comfortable with the vocabulary: what a load balancer, a cache, a message queue and a database replica are for. The course does not stop to define them, because it spends its time on design decisions instead. If those terms are new, take System Design Fundamentals first, which covers exactly that ground, and then come back here.

Text-based. Every lesson is a written design with diagrams, and there are more than 250 diagrams across the course. Text suits this material better than video: you can move at your own speed, jump straight back to a data model or an estimate without scrubbing a timeline, and copy API definitions into your own notes as you go.

In part. There is a full design of ChatGPT that covers the inference path, the model gateway, how generated tokens stream back to the client, and what happens when a connection drops mid-response. Several other problems cover the infrastructure that AI products lean on heavily, including the distributed cache, the key-value store, the rate limiter, and the metrics and monitoring pipeline.

Two weeks is enough to get broad and current coverage if you already know the fundamentals. What the plan really gives you is 30 worked designs plus the habit of comparing them: why WhatsApp routes through a session cache while Discord leans on a queue, why YouTube pulls content to the edge while Netflix pushes it there overnight. Interviews rarely ask a question you have memorised, so recognising which constraint forces which answer is the part that transfers. If your interview is further out, the same material rewards a slower second pass.

Yes. Design Gurus issues a certificate once you have completed the course, and you can share the link directly or add it to your LinkedIn profile.