Home Page
Schedule and Plans
Graphs
Internships and Resume section

Course Horse: AI-Powered Academic Planning for Penn State Students

Team USB_C - Alex Hasenbein, Preet Patel, Vivek Dadi, Rishika Thakre | HackPSU 2026

🎯 What It Does

Course Horse is an intelligent academic planning platform that helps Penn State students navigate course prerequisites, degree requirements, and semester scheduling. Instead of manually piecing together information from scattered PDFs and confusing prerequisite chains, students get:

Interactive Prerequisite Graphs: Visualize entire dependency trees for courses or majors
AI-Powered Schedule Generation: Automatically build semester-by-semester plans that respect credit caps and prerequisite chains
Horsey, the AI Academic Counselor: Chat with an AI assistant that understands Penn State's course catalog and answers questions
Real-Time Course Data: Integration with Penn State's course database for up-to-date information

The platform transforms academic planning from a frustrating guessing game into a clear, visual, data-driven process.

💡 Inspiration

Every semester, students face the same painful problem: "Which courses should I take next semester?"

The current process involves:

Digging through 200+ page PDF bulletins
Cross-referencing prerequisite requirements across multiple departments
Checking if courses are offered in fall/spring
Piecing together a 4-year plan without violating prerequisites
Hoping you didn't miss a hidden requirement

This is time-consuming, error-prone, and isolating since academic advisors are overbooked. We realized this is fundamentally a graph traversal problem—every course is a node, prerequisites are edges, and scheduling is a topological sort with constraints. But students shouldn't need to think in computer science terms. They should just get a clear answer: "Here's your optimal path to graduation."

🛠️ How We Built It

Architecture Overview

Course Horse is built as a modern full-stack web application:

Frontend (Next.js + React)
    ↓
API Routes (Node.js)
    ↓
MongoDB Atlas + Google Gemini 2.5

Tech Stack

Frontend:

Next.js 16 (React 19) - Server-side rendering and API routes
TypeScript - Type safety
Tailwind CSS 4 - Styling
React Flow - Interactive graph visualization
Zustand - State management

Backend:

Google Gemini 2.5 Flash - AI conversational agent
MongoDB Atlas - Course data and Penn State web content storage
Dagre - Graph layout algorithm
Custom scheduling algorithms - Semester plan optimization

Core Components

1. Prerequisite Graph Builder

The recursive graph builder traverses the dependency tree starting from a target course and builds visual nodes and edges representing courses and their requirements. It handles:

OR conditions (take A OR B)
Corequisites (must take together)
Cycle detection to prevent infinite loops
Depth limiting to avoid overwhelming graphs

2. Schedule Optimization Engine

The scheduler uses a topological sort algorithm to arrange courses in valid semester order, respecting:

Prerequisite dependencies (can't take a course before its prereqs)
Credit limits per semester (typically 18 credits max)
Semester availability (fall/spring offerings)

Our greedy packing approach prioritizes courses with many dependents to unblock future courses faster.

3. Horsey: The AI Academic Counselor

RAG Pipeline:

Extract course codes from user query (regex)
Search MongoDB for relevant course data and web pages
Fetch major requirements from bulletin
Build context and send to Gemini with conversation history
Gemini responds with text + optional graph/schedule actions

Key Features:

Hybrid search combining exact matches with semantic search
Multi-source fusion (course data + web pages + major requirements)
Structured actions to trigger graph/schedule visualizations
Context-aware responses based on what page the user is viewing

4. Interactive UI Components

Graph Visualization (React Flow):

Drag-and-drop nodes
Zoom and pan
Color-coded by completion status
Automatic layout via Dagre algorithm

Schedule Timeline:

Drag-and-drop course reordering
Live credit totals
Warning badges for overloads/conflicts

Chat Interface:

Markdown rendering
Context-aware (knows what page you're on)

📚 What We Learned

Technical Skills

Implementing graph algorithms (DFS/BFS) with cycle detection in production
Building effective RAG pipelines with proper context ranking
Designing AI agents with tool-calling capabilities
Real-time data visualization with React Flow
Balancing algorithm performance vs. user experience

Problem-Solving Insights

Start with a solid data model—type mismatches compound quickly
Build escape hatches when AI gets confused (manual overrides)
UX beats perfect algorithms—a good schedule shown instantly beats a perfect one that takes 10 seconds
TypeScript saved us from countless bugs during rapid development

🔮 What's Next for Course Horse

Short-Term

Elective recommendations using collaborative filtering
Multi-major and minor support
Export schedules to Google Calendar/iCal

Medium-Term

User accounts and saved plans
Track completed courses
"What-if" scenario planning (major switching)
Advisor integration for plan sharing

Long-Term

Expand to other universities
Career path integration showing typical outcomes
Predictive analytics for graduation risk
Mobile app with push notifications

🏆 Conclusion

Course Horse transforms academic planning from a frustrating manual process into an intuitive, visual, AI-assisted experience. By combining graph algorithms, AI-powered chat, and real-time course data, we've built a tool that helps students save time, reduce stress, make informed decisions, and graduate on time.

Beyond the technical achievements, Course Horse represents a shift in how students interact with university bureaucracy—making it visual, interactive, and conversational instead of buried in PDFs.

🔗 Links & Resources

GitHub: USB_C_HackPSU-course-horse-frontend & USB_C_HackPSU-WorkingAgent
Tech Stack: Next.js, React, TypeScript, Google Gemini 2.5, MongoDB, React Flow, Dagre
Team: USB_C - Alex Hasenbein, Preet Patel, Vivek Dadi, Rishika Thakre

Built with ❤️ at HackPSU 2026