Case study · Web
Admission AI Chatbot
A retrieval-augmented chatbot that answers admission questions from a college's own data.
- Role
- Full-Stack Developer (Freelance)
- Platform
- Web
- Timeline
- 2024
- Stack
- Python, Flask, LangChain
01 — Context
Prospective students ask the same questions again and again — deadlines, eligibility, fees, courses — and the answers are scattered across college documents.
The goal was a chatbot that answers from that data rather than from a language model's general knowledge, so students get the college's answer, not a plausible-sounding one.
02 — What I built
Grounded in college data
Retrieval over the college's own material so answers come from the source.
Two-service architecture
A Python Flask service runs the LangChain pipeline with OpenAI; a separate Next.js app serves the chat UI.
Google sign-in
Authentication with Google via NextAuth.
03 — How it's built
- Keeping the AI service in Python — where the LangChain ecosystem is strongest — and the product UI in Next.js let each side use its best tooling and deploy independently.
- LangChain orchestrates retrieval and the OpenAI calls behind a small Flask API.
- Python
- Flask
- LangChain
- OpenAI
- Next.js
- NextAuth