RAG Academic Advising Agent
An AI agent designed to automate complex academic advising inquiries at UIUC.
Developed a Retrieval-Augmented Generation (RAG) prototype for the ATLAS Exploration Team at UIUC. The agent serves as an automated academic advisor, capable of answering complex degree requirements and logistical questions using high-precision semantic search.
System architecture pipeline: From PDF document ingestion and text chunking, to vector semantic search, and final LLM response generation with citations.
Core mechanisms: Navigating the knowledge base, filtering for relevance, and quoting sources accurately to eliminate hallucinations.
Key Technical Implementations
- Vector Database Pipeline: Engineered a robust pipeline to ingest and index unstructured administrative data (PDF handbooks, FAQs), enabling high-precision semantic search.
- Precision Semantic Search: Implemented capabilities to navigate complex degree requirement hierarchies that traditional keyword search might miss.
- Prompt Engineering & Evaluation: Developed specialized system prompts and quantitative evaluation metrics to minimize hallucinations, benchmarking accuracy against human advisor standards.
- Automated Logistics: Designed the agent to handle repetitive student inquiries using Python, significantly reducing manual workload.
Note: This project was developed as part of the ATLAS Exploration Team at the University of Illinois Urbana-Champaign.