Open source · system design academy
The system design interviewer that actually interviews you.
174 lessons from first principles, then a live mock interview that never gives you the answer, catches your contradictions and scores you on 8 dimensions.
01 / The interview
It never gives you the answer.
It asks the question that reveals whether you know it. Name-drop a technology and it asks why. Contradict yourself and it notices.
An illustration of the engine's behaviour; the interviewer's lines are from the project README.
02 / The curriculum
From "what is a computer" to ML systems.
Seven tiers, 41 modules, 174 lessons. Every point of light beside this text is one real lesson, floor by floor. Explore them:
The Foundations
4 modules · 21 lessons
03 / The engine
Only the voice is an LLM.
Four layers. Three of them are deterministic code, so the interviewer always catches contradictions, always probes shallow claims and always moves on at the right time, whatever the model. You can't talk it into a better grade.
- 4SpeakerTurns actions into natural dialogueLLM
- 3StrategistPicks the next move from 11 priority tiersCode
- 2TrackerCoverage, contradictions, depth, phase, rubricCode
- 1AnalyzerConcepts, numbers and depth from your wordsCode
04 / The scorecard
Not pass or fail. Exactly where you're weak.
Every interview ends with a rubric across eight dimensions, against the bar for the level you chose.
Scorecard · Staff bar 75
- Requirements gathering8/10
- Capacity estimation10/10
- API design6/10
- Database design8/10
- Caching strategy8/10
- Scalability10/10
- Failure handling8/10
- Communication8/10
82/100Passed
Example scorecard, from the project README.
05 / The interviewers
Warm for SDE2. Unforgiving for Staff.
Five personas, each with its own time limit, tone and bar.
The Principal Engineer
Direct, no hand-holding · 45 min
Drive everything. Go two or three levels deep. Production-grade failure handling.
06 / Run it
One command. No API key.
# prebuilt images from GitHub Packages$ git clone https://github.com/Ayush1906saxena/architectiq.git$ cd architectiq$ docker compose pull && docker compose up# open http://localhost:3000The interviewer runs on the bundled Ollama (llama3.2, pulled on first start). Add a free Groq key for faster answers.
- ghcr.io/ayush1906saxena/architectiq-frontend
- ghcr.io/ayush1906saxena/architectiq-backend
Built by Ayush Saxena · Next.js, FastAPI, Groq or Ollama · this page: three.js TSL compute