Clarift
AI-powered study engine for Filipino students — transforms uploaded materials into summaries, quizzes, and targeted practice
Problem
Filipino students and review center learners waste hours reformatting raw study material into usable review content. Existing tools either ignore uploaded notes or generate content from general knowledge rather than the student's own materials.
Solution
Clarift is an AI-powered study engine built specifically for Filipino students and review center learners. It transforms uploaded study material into structured summaries, quizzes, and targeted practice through an active learning loop designed for high-stakes exams. The system uses RAG to ground all AI outputs in the student's own notes, never generating from general knowledge.
Role
Sole Full Stack Developer
Stack
- Next.js 16
- React 19
- TypeScript
- Tailwind CSS 4
- shadcn/ui
- Radix UI
- Framer Motion
- TanStack React Query
- Zustand
- Tiptap
- Recharts
- FastAPI
- Python 3.12
- LangChain
- SQLAlchemy async
- Alembic
- Neon PostgreSQL
- pgvector
Integrations & Services
- Clerk
- Google Gemini API
- Cloudflare R2
- Upstash Redis
- PayMongo
- Sentry
- Vercel
- Railway
Key Features
- 01
Document ingestion — upload PDFs as study material
- 02
Structured summaries — multi-step AI chain with MermaidJS diagrams
- 03
Quiz generation — auto-generated from uploaded material only
- 04
Weak area diagnosis — identifies knowledge gaps from quiz performance
- 05
Targeted practice — personalized drills focused on specific weak topics
- 06
Grounded RAG chat — answers exclusively from uploaded notes
- 07
Quota system — daily usage limits with free and Pro tiers
03 / Technical Decisions
- 01 /
Split architecture (Next.js frontend + FastAPI backend): AI pipelines need persistent processes with proper task isolation, serverless functions hit timeout limits for LangChain chains.
- 02 /
ARQ worker with Redis queue: async job processing for document ingestion and AI generation keeps the API responsive and allows SSE progress tracking.
- 03 /
pgvector for embeddings: leverages the existing Neon PostgreSQL database rather than adding a separate vector store, reducing infrastructure complexity.
- 04 /
Clerk for auth with JWT verification in FastAPI: OAuth handled by a dedicated service, backend verifies tokens directly for tenant isolation.
- 05 /
Cloudflare R2 for file storage: S3-compatible, cost-effective for PDF uploads without vendor lock-in.
Screenshots
