§ Case 06 — Product · AI
CrackIt2027
Mobile exam-prep companion with a custom RAG pipeline, syllabus-aware quizzes, and a quiet, paper-like interface. 2027 update adds an offline mode and live tutor handoff.
Year
2027
Role
Product Design
Stack
React Native · LangChain
Award
Product Hunt · #3 of the Day
Case 06 · 2027
Product · AI
01 · Brief
CrackIt — brief
University-entrance prep in South Asia is dominated by 4 000-page textbooks and last-minute coaching. CrackIt's premise: the same content, delivered as a quiet, paper-like companion that knows your syllabus and your weakest topics.
02 · Approach
Approach
A custom RAG pipeline indexes the ten official textbooks per board and retrieves passage-level citations on every answer. Quizzes are generated from the same passages so users can always trace back to the page. The 2027 update adds an offline mode — the most-used 12 % of content fits in 80 MB on-device.
03 · Solution
Solution
The interface is paper-like — single-column, generous leading, no notification badges. Streaks are present but small. The 'live tutor' button sends a snapshot of your last 30 minutes to a human reviewer if the model can't answer with confidence above 0.78.
◊ Gallery
Selected frames
Click any image to enter the lightbox. Use ←/→ to step through, Esc to close.
◊ Numbers
180k
Monthly active
22%
Paid conversion
+14 pp
Avg. score lift
80 MB
Offline corpus
◊ Deliverables
- 01React Native app (iOS + Android, single codebase)
- 02RAG pipeline (LangChain + pgvector + reranker)
- 03Syllabus-aware quiz generator with passage citations
- 04Offline-first sync engine (delta-synced 80 MB corpus)
- 05Live-tutor handoff workflow + reviewer dashboard
- 06Quiet design system: paper backgrounds, single-column reading
04 · Outcome
Outcome
180 000 monthly active users, 22 % paid conversion, +14 percentage-point average score lift on mock exams. Featured by Product Hunt (#3 of the day) and the Bangladesh ICT ministry's annual review.
My daughter prepares with this for two hours a day and never asks me to charge the phone. That's the highest review I have.
◊ Stack
React Native
LangChain
Supabase