
Move a meeting scheduler's bookings out of process memory into the service's database through migrations, so bookings survive deploys, both instances agree, times are stored in UTC and concurrent requests for one slot can never double-book a host; then give bookings a length with a migration the still-running previous release can live with.
Fix group registration in an event ticketing service so the order, attendees, ticket codes, confirmation email and seat count change together or not at all, the last seats are never oversold, cancelled events are refused, and a slow ticket service never holds up other buyers.
Get a notes app ready to run as one build in two regions: read its settings from environment variables, refuse to start with one error naming every missing setting, and add liveness and readiness probes that check the database and storage, answer within two seconds and never leak connection details.
Build the pickup service of a parcel locker network: migrated lockers, compartments and parcels, deposits that atomically take the smallest free compartment and queue the pickup SMS, code-checked pickups, a schema change the previous release survives, configuration from the environment and health probes.
Design backend services that survive real traffic and real failures. Starting from a service that keeps its state in a database, configures itself from the environment and reports its health, you add the safeguards production systems rely on: rate limits, caching, idempotency, queues, resilience patterns, load balancing and observability.
This path is system design you write, not draw. Each challenge hands you a working backend service, such as a meeting scheduler, an event ticketing service or a maps API, and asks you to make it hold up in production. You work in Java or Python. The first module, Service Foundations, moves state out of process memory into a migrated database. It makes multi-step writes atomic, so the last seat is never sold twice, and lets one build run anywhere with environment configuration and liveness and readiness probes. The second, Rate Limiting, protects APIs from abusive clients with fixed-window, sliding-window and token-bucket limits, per key, per user and per IP, with standard limit headers and limiter state shared across instances. All 8 challenges are medium difficulty. They suit developers who can already build a small web service and want practice with the problems that appear once it runs on more than one instance. More modules are planned. On every challenge you choose whether to work with AI agents, AI chat only, or no AI.
2 modules: Service Foundations and Rate Limiting.
8 exercises, all medium.
You can work in Java 25 or Python 3.13.
No. Every challenge runs in a browser-based IDE with everything already set up, and your code is graded automatically against a test suite.
It’s hands-on, not whiteboard. You change a real service’s code and tests check that it holds up under concurrent requests, restarts and several instances. Rate limiting and consistent writes come up often in system design interviews, so the practice carries over.
You should be able to build a small web service in Java or Python and write basic SQL. No earlier system design experience is needed.
Yes. Both modules are open on a free account, within the free plan’s weekly hour and monthly solution unlock.
Yes, if you want to. Each challenge lets you choose how to work: with AI agents, with AI chat only, or with no AI at all.