LionPlanner

No more advisors no more headaches!

Comment

Inspiration

Every Penn State student knows the pain of LionPATH. Planning a semester usually means juggling five different tabs. This may include a messy advising transcript, a confusing degree bulletin, RateMyProfessors, and a blank spreadsheet. We wanted to build something better than just another generic calendar app. We wanted to build a true AI academic advisor that actually understands your major, your past grades, and your lifestyle to build the perfect schedule.

What it does

LionPlanner is an AI powered academic planning engine that turns your PSU advising transcript (PDF) into a tool where you can browse next semester courses and what you currently need.

  • Stateless Degree Audit: Drop your advising transcript PDF into the app, select your major, and our backend dynamically fetches the official Penn State bulletin to calculate exactly what classes you still need.

How we built it

  • Frontend: Next.js 16 App Router, styled with Tailwind CSS, Framer Motion, and Aceternity UI
  • Backend: Python FastAPI
  • The AI Engine: We used the Google Gemini 2.5 Flash API. Instead of storing massive databases of degree requirements, we built a pipeline that dynamically fetches the official PSU Bulletin PDF into memory, feeds it to Gemini alongside your transcript, and forces a strict JSON output of your missing classes.
  • Data Pipeline: We used Playwright to bypass Oracle PeopleSoft and scrape live course sections, and built a custom backtracking constraint solver to generate over 200 schedule candidates before scoring them.

Challenges we ran into

Our biggest hurdle was data acquisition. Oracle PeopleSoft is notoriously hostile to scrapers, requiring us to build highly resilient Playwright scripts to navigate hidden iframes and AJAX spinners. Also we felt like feeding a 10 page degree bulletin and a transcript into an LLM initially caused hallucinations. We solved this by enforcing strict Pydantic schemas and building a dictionary that instantly routes any major to its exact source-of-truth PDF.

Accomplishments that we're proud of

  • Building a 100% degree audit engine that works for any major without needing a database.
  • Successfully merging a deterministic algorithm with gen AI logic.

What we learned

We leveled up our full stack architecture skills, specifically learning how to properly wire Next.js to FastAPI without CORS headaches. We also learned advanced prompt engineering (forcing an LLM to output perfectly structured JSON arrays to drive complex UI components.)

What's next

  • Adding 1-Click .ics Calendar Export
  • Integrating live seat availability
  • Scaling the dictionary to support universities beyond Penn State
Share this project:

Updates