Syllabus-Sync

01 / Overview

Problem

Australian universities publish syllabi as unstructured PDFs. Students manually copy assessment dates into calendars, and AI assistants hallucinate deadlines because no machine-readable source of truth exists.

Solution Overview

1. Built an LLM OCR pipeline that extracts structured JSON from PDF syllabi with Zod schema validation
2. Implemented Syllabus-as-Code where assessments and deadlines are version-controlled, diff-able artefacts
3. Created a full student OS with calendar, campus navigation (Leaflet + Google Maps), and push notifications
4. Designed for other Australian universities to adopt by replacing data/ and setting environment variables

02 / Evidence

Proof & Verification

• Accepted into the Macquarie University Incubator as a formal startup (2026)
• Live in production at syllabus-sync.app
• All 503 tests across 92 files passed in CI
• Zero-Trust proxy catches misconfiguration by default (validated in production)
• Designed to serve ~47,000 MQ students; architected to fork for other Australian universities
• WebAuthn passkey authentication keeps shared secrets on the device

03 / Methods

Build

Tech Stack

• Next.js 16
• TypeScript 5
• Supabase PostgreSQL
• TanStack Query
• Leaflet + Google Maps
• Framer Motion
• Radix UI
• Sentry

• LLM OCR pipeline extracting structured JSON from PDF syllabi
• Syllabus-as-Code: assessments and deadlines are version-controlled artefacts
• Full student OS: calendar, reminders, campus map, notifications
• 503 tests across 92 files with CI/CD quality gates

Secure

• Zero-Trust proxy middleware authenticates every route by default
• Supabase Row-Level Security enforced at database layer
• WebAuthn/FIDO2 passkey authentication (no passwords stored)
• LLM prompt injection mitigation with schema-constrained output and input sanitisation
• Secrets scanner in CI blocks credential patterns from commits