All projects

Case study · AI & ML

Lyon: NBS Degree Advisor

An agentic degree advisor for Nanyang Business School that matches candidates to graduate programmes and answers their questions around the clock.

Context
Course group project, AN6001 AI and Big Data, Nanyang Technological University
Stack
  • GPT
  • LangChain
  • RAG
  • pgvector
  • React
  • FastAPI
  • Supabase
  • Vercel
  • Python
Programmes covered
11
Document chunks indexed
1,400+
Agent tools
5

The problem

Prospective students compare 11 graduate programmes whose details are spread across brochures, web pages and admissions emails.

Admissions staff answer the same questions repeatedly, and candidates who are unsure where they fit have no structured way to find out.

Data

Programme pages and admissions content split into 1,400+ text chunks, embedded and stored in pgvector on Supabase.

Candidate input from an uploaded CV and a short profiling quiz.

Approach

  1. 01

    Knowledge base

    Programme and admissions content chunked, embedded and indexed for retrieval.

  2. 02

    Agent

    A GPT model orchestrated with LangChain chooses between 5 tools, including knowledge search, programme comparison, FAQ lookup and handoff to a human advisor.

  3. 03

    Candidate profiling

    The CV and quiz answers are scored on 7 competency axes and matched to programmes, with the fit shown on a spider chart.

  4. 04

    Delivery

    A React front end and FastAPI back end, deployed on Vercel, with Lyon presented as the NTU lion mascot.

Impact

Candidates can compare programmes side by side and get answers on admissions, fees and deadlines at any hour, with a route to a human advisor when needed.

Next steps

  • Evaluate answer quality against a labelled question set and track retrieval accuracy.
  • Show source citations with every answer.
  • Log unanswered questions to guide updates to the knowledge base.