ChronosAI

Snap your homework and never miss a deadline again! ChronoAI uses AI to transforms photos of homework into organized, optimized study time on your Google Calendar in seconds.

Comment

Inspiration

We’ve all felt that sinking moment: realizing the night before that you forgot about a test, a major project, or an assignment buried in a long list of commitments. As two students navigating the heavy workload of junior year, we constantly felt overwhelmed by assignments, deadlines, and extracurriculars. Surprisingly, the hardest part wasn’t completing the work, but managing the limited time between all our different commitments. We realized that while many students write down homework, few actually convert it into structured time. Google Calendar was a good option, but inputting every task manually was always a hassle. ChronoAI was born from our frustration with procrastination, missed deadlines, and poorly planned study time.

What it does

ChronoAI turns a photo of a homework list into a real study plan on Google Calendar. A student uploads a planner/whiteboard/LMS photo, the app extracts assignments, estimates effort, lets the student quickly edit anything wrong, then creates conflict-aware study blocks before due dates.

How we built it

  • Frontend: Next.js + React + TypeScript + Tailwind
  • Auth: Google OAuth via NextAuth
  • AI extraction: Azure OpenAI vision model for OCR + structured task parsing
  • Scheduling: custom rule-based scheduler (due-date aware, block splitting, cutoff time, conflict avoidance)
  • Calendar integration: Google Calendar API (read existing events + create study blocks)
  • Deployment: Vercel

Challenges we ran into

  • Google OAuth setup issues (redirect_uri_mismatch, test users, consent-screen mode)
  • Vercel environment configuration mismatches between local and production
  • Timezone bugs (events landing at 1 AM due to server timezone differences)
  • Preventing scheduling in past times for same-day tasks
  • Making model output reliably parse into strict JSON from vision responses

Accomplishments that we're proud of

  • End-to-end flow works: photo -> extracted tasks -> editable list -> scheduled calendar events
  • Fast, demo-friendly UX with a correction step to handle OCR/LLM uncertainty
  • Timezone-aware scheduling fixed for Bay Area/local user context
  • Stable Google Calendar write integration with consistent event formatting

What we learned

  • OAuth is often the hardest non-ML part of shipping AI apps
  • Timezone handling must be explicit in both scheduling logic and event creation
  • AI extraction quality improves a lot when users can quickly edit outputs
  • MVP success comes from reliable system design, not just model quality

What's next for ChronoAI

  • Handle all-day calendar events and deeper conflict logic
  • Smarter prioritization (difficulty, class weight, urgency)
  • Auto-rescheduling when a block is missed
  • Notifications/reminders before each study block
  • Multi-user onboarding and polished production launch (Google verification + analytics)

Built With

Share this project:

Updates