Retriever AI

The future of academic advising isn't human vs. machine - it's human enhanced by intelligent insights. And that future starts at UMBC, today.

Comment

Inspiration

we were provided with a comprehensive Neo4j student knowledge graph containing academic relationships, course data, and student information, and tasked with creating something impactful from it. Recognizing the untapped potential in this rich educational dataset, we were inspired to build a comprehensive academic success platform that could transform how students navigate their educational journey using the power of graph analytics and machine learning.

What it does

UMBC Academic Analytics Platform is a comprehensive student success solution built on top of the provided Neo4j knowledge graph. We transformed the raw graph data into a powerful platform featuring:

  • Academic Risk Prediction: Machine learning models that predict student success in courses by analyzing graph relationships, learning styles, and academic patterns
  • Intelligent Course Planning: AI-powered course recommendations based on prerequisite chains, similar student success patterns, and optimal degree pathways
  • Study Group Matching: Smart compatibility system that connects students with similar learning styles and academic goals
  • Progress Tracking Dashboard: Visual analytics showing academic progress, course completion patterns, and achievement tracking
  • AI Academic Advisor: Chatbot that queries the knowledge graph to answer student questions and provide personalized guidance
  • Mentorship Network: Intelligent matching system connecting students with peers, faculty, and industry professionals based on graph relationships

The platform leverages the provided knowledge graph's 500+ students, 100+ courses, faculty relationships, and complex academic networks to deliver evidence-based recommendations.

How we built it

Starting Point: Provided Neo4j Knowledge Graph: Rich dataset with students, courses, faculty, textbooks, prerequisites, and academic relationships

Graph Analytics: Extracted insights using Cypher queries to understand academic patterns and student similarities

Backend Development:

  • FastAPI Server: Built comprehensive REST API with 20+ endpoints serving different platform features
  • Graph Query Optimization: Developed efficient Cypher queries for complex academic relationship traversals
  • Machine Learning Pipeline: Training models on graph-derived features required careful feature engineering to convert graph relationships into meaningful numerical representations ## Challenges we ran into
  • Neo4j Learning Curve: As newcomers to Neo4j, we struggled with Cypher query syntax and graph database concepts
  • Data Loading Issues: Importing CSV data into Neo4j was time-consuming and required multiple attempts to configure relationships properly
  • Limited ML Training Data: With only 500 students and 100 courses, traditional ML algorithms had insufficient data, so we pivoted to similarity-based approaches

Accomplishments that we're proud of

  • Successfully built a full-stack application that transforms raw graph data into actionable insights for students, creating a seamless experience from login to mentorship matching
  • Delivered six major features (risk prediction, course planning, study groups, progress tracking, AI advisor, and mentorship) that work together to support student's throughout their academic journey.
  • Built everything using actual student data and relationships from the provided knowledge graph, avoiding hardcoded assumptions and creating genuinely useful recommendations.

What we learned

  • Working with real-world data taught us the importance of robust data cleaning and validation - inconsistent naming conventions and missing relationships can break entire features
  • Neo4j's ability to traverse complex relationships opened our eyes to how academic data can reveal hidden patterns and connections that traditional databases miss
  • Designing FastAPI endpoints that serve both simple queries and complex graph traversals requires careful performance consideration

What's next for Retriever AI

  • Develop native mobile apps for iOS and Android to make academic planning accessible anywhere, with push notifications for important academic milestones
  • Partner with UMBC's existing student information systems to provide real-time data synchronization and automated academic planning
  • Optimize the platform to handle larger datasets and more concurrent users, potentially migrating to cloud infrastructure for better performance and reliability
  • Extend the knowledge graph to include industry connections, internship opportunities, and career outcomes to help students make informed decisions about their academic and professional futures
Share this project:

Updates