Beta · For Software, Data and AI Engineers
You learn system design by designing.
You draw the architecture on a canvas, we grade it instantly with rules that explain every point, and you come back tomorrow for the streak. Like learning a language: a little, every day.
No card needed · the placement test uses no AI and takes about 10 minutes
Application LB
ECS / EKS
ElastiCache
RDS / Aurora- The client reaches a service+8
- A load balancer sits at the entry+6
- A cache absorbs the reads+4
- The client talks straight to the database0
- load balancer
- cache-aside
- consistent hashing
- idempotency
- fan-out
- sharding
- backpressure
- outbox
- quorum
- rate limiting
- CDN
- read replicas
- CQRS
- geohash
How it works
From placement test to your first graded design, in under 15 minutes.
The same order as a real interview: first you scope, then you estimate, and only then you draw.
01 · We place you in 10 minutes
12 quick questions, no AI. Your path starts at the level you show.
02 · You design in 5 steps
Requirements, estimates, diagram, deep dive and trade-offs.
03 · Immediate grading
Fixed rules set the score. In Pro, you also get interviewer feedback.
Product
Built for practice, not for reading.
Application LBECS / EKSElastiCacheRDS / AuroraA canvas that understands systems
Every component has a role. That's how we know the client shouldn't talk to the database, even if you named it “my-turbo-db”.
ElastiCache
Memorystore
Azure Cache for RedisAbstract, AWS, GCP or Azure
Switch clouds and the diagram adapts. The score stays the same.
monolith → cache → checkout → … → multi-region
A system that grows with you
Each stage starts from your previous diagram and adds a requirement that breaks it.
Streak, XP and units
15 minutes a day beat a cram session right before the interview.
In English and Spanish
Practice in your language and switch to English as the interview gets closer.
challenges
0
campaigns
0
units
0
canvas modes
0
From estimating Twitter's QPS to a Databricks lakehouse or a RAG with its evaluation. See the catalog
Pricing
Less than a one-hour mock with a person.
Regional pricing in LatAm: up to 50% off. Prices in USD, to be confirmed in the beta.
Free
USD 0
To start the habit.
- 5 challenges from your roadmap, junior to staff, and the daily case
- Streak, XP and placement test
- 1 hint per challenge
- Rule-based grading
- 1 trial of interviewer feedback
Pro
recommendedUSD 6.58
per month · USD 79 billed yearly
- Every challenge and the 22 campaigns
- All 3 hints of every challenge
- AI interviewer feedback
- All three roadmaps: Software, Data and AI Engineer
Lifetime launch deal
USD 149
One payment. For the first 100 on the waitlist.
- Everything in Pro, forever
- New features too, like mock interviews
Cuby for teams
System design interviews that measure what matters.
For teams that hire: the candidate designs on Cuby's canvas, then builds part of it with Claude Code, live. You see how they think, decide and work with AI, and you get a report you can compare.
Live design
Watch the architecture as the candidate draws it, graded by the same rules that check every objective.
Live coding with Claude Code
They build part of the design with Claude Code alongside. You assess what they ask it, what they review and what they discard.
Every session is recorded
Each stroke of the diagram and each command stay on a timeline to go over with your team.
A report you can compare
Scores per objective, trade-offs and seniority signals, with the same yardstick for every candidate.
API Gateway
API GW throttlingECS / EKSElastiCacheClaude Code
> add a per-user rate limiter to the API
- ● Reading api/handlers.go
- ● Editing middleware/ratelimit.go +42 −3
- ✓ Tests: 18 passed
Evaluation
- Scalability
- Trade-offs
- Use of AI
FAQ
Questions people ask us.
Yes. The placement test passes the units you already know, with half the XP, and your path starts with what you need to reinforce. Campaigns for seniors start from the advanced stages.
No. Abstract mode is the one most interviews use. The cloud modes are optional, and your score is the same in all four.
No. Your score comes from fixed rules applied to your diagram and your answers, so it is fair and repeatable. AI only comments: what was missing, which trade-offs to discuss and what follow-up questions you would get.
Not entirely, and it doesn't promise to. It gets you to the mock with a well-oiled method, so your hour with a person goes to what matters.
All content is in English and Spanish. You can switch languages anytime, even in the middle of a challenge.
Your next interview starts with 15 minutes today.
Open beta in November
The first 100 get the lifetime launch deal.