Case study · Salil Lakra · 2026
Quill
Campus operations and assessment on a shared document engine.
Solo full-stack
Problem
Campus work is split across spreadsheets, PDFs, and ad-hoc tools. A timetable, a syllabus, and a marking stack rarely share one document engine.
What I built
Quill is a workspace where a PDF becomes the source of truth. After ingest, you can ask the document with citations, extract a timetable into campus ops, generate a question paper, and grade scripts with a rubric.
Architecture
Next.js 16 on the client. Bun + Hono API, Prisma on PostgreSQL, Redis/BullMQ workers for ingest, extract, generate, and grade. Pinecone for vector search, optional Gemini, Socket.IO for live pipeline progress. Demo mode runs without an API key.
My role
Solo author. License © 2026 Salil Lakra. Backend API, four BullMQ workers, Prisma schema, role-based Next.js app, Docker Compose, and GitHub Actions CI.
Frames
