DEPSTAR Admission Bot
University Inquiry & Guidance RAG System
The Problem
University admission periods overwhelm administrative staff with repetitive, documentation-heavy queries, frustrating prospective students.
Why it matters
Rapid, accurate response times directly correlate with higher enrollment rates and student satisfaction.
Who is affected
Prospective students, parents, and University Admins.
Architecture & System Design
Frontend
Sleek, chat-focused interface built with React & Next.js for zero-latency interactions.
Backend Processing
LangChain orchestrating embeddings retrieval from a Pinecone vector index before query generation.
ML Pipeline
Advanced RAG pipeline combining BM25 keyword search with dense vector semantic search for ultimate precision.
Architectural Reasoning
A dual-retriever approach (hybrid search) ensures specific acronyms are caught by BM25, while contextual questions are handled by dense vectors.
Alternatives Considered
A standard dense-only retriever failed to accurately fetch highly specific course codes.
Core Features
Hybrid Search RAG
Fuses semantic intent with exact keyword matching.
Source Citation
Transparently links students to the official university document referencing the answer.
Agentic Guardrails
Strictly refuses to answer non-university queries to prevent PR liabilities.
Results & Impact
Migrated a legacy FAQ page to an interactive, conversational AI system.
Currently serves as the first point of contact for DEPSTAR university prospectives.
10k+ Queries Handled
0% Hallucination on Core Policies
Takeaways & Learnings
What I Learned
Learned how to construct iron-clad System Prompts to prevent jailbreaks in a public institutional setting.
Trade-Offs Made
Prioritized strict factual grounding over conversational flair.
Future Improvements
Integrate multi-lingual support for international student queries.
Tech Stack Foundation
ML / AI
- LangChain
- OpenAI Embedding
- Hybrid RAG
Frontend
- Next.js
- TailwindCSS
Tools
- Pinecone
- Vercel
Interested in this architecture?
Let's talk about how I can build something similar for your team.