Overview
The SMU AI Resume Generator is a commercial AI content-generation platform built for a Singapore university and its career-services program. It serves roughly 13,000 students, letting them generate polished resumes and career content on demand. Rather than producing generic boilerplate, it grounds every draft in the student's own profile and the institution's career guidelines.
The Problem
Students face a slow, uneven path from a blank page to a resume that reflects both their experience and their school's standards, and career-services teams cannot review every draft by hand. The platform replaces that manual back-and-forth with guided, on-demand generation.
Key Features
- RAG-grounded output — a Retrieval-Augmented Generation agent pulls from each student's profile and the institution's guidelines so generated content reflects real experience rather than invented filler.
- Tool-calling agent — the agent invokes tools to gather and structure the information it needs, assembling career content step by step instead of producing a single unguided completion.
- Dynamic prompting — a dynamic-prompt system tailors each generation to the individual student and the university's requirements, keeping output specific and on-standard.
- On-demand generation — students generate resumes and other career content whenever they need it, turning a blank page into a structured, review-ready draft in a single session.
- Institution-aligned content — every draft is shaped by the career-services program's own guidelines, so output stays consistent with what the university expects from its students.
- Campus-scale platform — the system is built to serve the university's full student population of roughly 13,000, delivering personalized generation across the cohort.
How It's Built
The platform is built on Next.js for the student-facing application, backed by a Node.js service layer that runs the tool-calling RAG agent and dynamic-prompt system. PostgreSQL stores the student profiles, institutional guidelines, and generated content that ground the agent's output. The retrieval layer feeds each student's profile and the school's guidelines into generation so results stay personalized and aligned with career-services standards.
Tech Stack
Note
Client project; the source is private, so this write-up stays at a high level.