Coding Practice on Real Broken Systems
Most coding practice is toy problems in a sandbox. This is the real thing: you open a broken system in a live cloud workspace and fix it - an API, a query, an outage - the way the job actually works. Each fix is saved to a portfolio.
Why practice on real systems
Toy problems build narrow skills that do not transfer to a messy production codebase. Real systems build the skill that does: reading unfamiliar code, forming a hypothesis, and fixing it under realistic constraints.
18 challenges to practice
Build an LLM Content Moderation ServiceAI/ML · MidCut OpenAI Costs by Caching With RedisAI/ML · MidPower RAG Search With a Vector DatabaseAI/ML · MidBuild a Webhook Delivery System With RetriesBackend · MidFix the N+1 QueryBackend · MidStop a Double-Charge Race ConditionBackend · MidBuild a Data Quality Monitoring APIData · MidSchedule a Pipeline as an Airflow DAGData · MidBuild a Staging-to-Marts dbt ProjectData · MidTrace an Nginx 503 CascadeDevOps · MidPlug a Postgres Connection Pool LeakDevOps · MidIndex a Slow Postgres Search QueryDevOps · MidMigrate a React App From JavaScript to TypeScriptFullstack · MidWire React to Supabase (CRUD + Realtime)Fullstack · MidAdd a JWT Login and a Protected Route in ReactFullstack · MidWrite a PodDisruptionBudget and Spread ConstraintsKubernetes · MidWrite an HPA That Scales on a Custom MetricKubernetes · MidConfigure a Kubernetes HPA for AutoscalingKubernetes · Mid
How to get the most from it
Practice a little across different areas rather than grinding one. Breadth is what makes you adaptable, and each finished fix is a concrete portfolio piece you can point to.
Build your portfolio free. You are in a real cloud workspace in 30 seconds - fix real systems, show real work.
Start free →