Inspiration
In many Kenyan schools, learners share devices and data is expensive. Teachers already have syllabi in PDFs/Docs—but turning them into bite-size lessons, quizzes, and voice notes takes time. ClassKit AI converts a syllabus into a WhatsApp-first micro-course that works on low data, bilingual (English/Swahili), and offline-leaning delivery.
What it does
Uploads a syllabus (PDF/Doc) and auto-segments it into lessons.
Generates summaries, 5-question quizzes, and voice notes per lesson.
Learners interact via WhatsApp (or web) with progress tracking and gentle nudges.
Teacher mode: edit lessons, download printable sheets, view class insights.
How we built it
Ingestion: parse PDFs/Docs → chunk by headings → clean tables → store in Postgres.
Lesson engine: semantic chunking + prompt templates → summaries, quiz Q/A, TTS script.
Delivery: WhatsApp webhook → session state → rate-limits + retry queue.
Analytics: per-student mastery score ( M = \frac{c + 0.5r}{n} ) where c = correct, r = review completions, n = total items.
Mini architecture
[Upload/API] -> [Parser + Chunker] -> [Vector DB] -> [Lesson Generator] -> [Quiz + TTS] [WhatsApp Bot/Web] <-> [Session + Progress Store] <-> [Analytics Dashboard]
Challenges
Low-data UX: keeping WhatsApp messages concise without losing pedagogy.
Multilingual prompts: swapping between EN/Kiswahili while preserving accuracy.
PDF chaos: tables, headers/footers, and non-selectable text.
Accomplishments we’re proud of
<60s from syllabus upload to a first draft micro-course.
Seamless WhatsApp flow with lesson paging and quick-reply buttons.
Teacher dashboard that surfaces “what to reteach” automatically.
What we learned
Better outcomes when quizzes use worked examples before free recall.
TTS speed at ~0.9× and sentence-level pauses improved comprehension for Form 1–2.
Log in or sign up for Devpost to join the conversation.