Career Compass: an AI career advisor
Files
TR Number
Date
2025-05-05
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Career Compass is an AI-driven career guidance platform that combines Retrieval-Augmented Generation (RAG), structured labor data from O*NET and BLS, and large language models (LLaMA 3.3) to deliver personalized, explainable career insights. The platform offers a real-time conversational chatbot and an embedded Career Insights Dashboard built in Tableau, enabling users to explore occupations, required skills, education, and job outlooks interactively. Designed for accessibility and scalability, it includes a semantic vector search engine, persistent conversational memory, and structured fallback logic to ensure accurate responses. The data visualization component translates complex employment and wage statistics into narrative-driven dashboards and interactive data stories, making labor trends more understandable for users without technical backgrounds. The application is deployed on AWS EC2 with a Next.js frontend and FastAPI backend.
Description
Career Compass is an AI-powered career exploration platform that combines large language models, structured occupational data, and interactive data visualizations to support students and professionals in making informed career decisions. Developed by Group 2 as part of Virginia Tech’s CS 5934 Capstone Project (Spring 2025) guided by Prof. Sara Hooshangi, the platform offers an intelligent chatbot experience, career dashboards, and seamless end-to-end integration deployed on AWS.
Backend:
The core engine is a Retrieval-Augmented Generation (RAG) pipeline built with FastAPI and powered by OpenAI's `text-embedding-ada-002` model and the ChromaDB vector store. It processes over 30 structured files from O*NET 29.1 (~50MB), including job descriptions, skills, tools, and education levels. Data is cleaned, standardized, and restructured into markdown-style formats for efficient LLM prompting. The backend supports semantic vector search, dynamic prompt construction, and session-based memory using LangChain, interfacing with Meta’s LLaMA 3.3 (70B) hosted through the Groq API. Guardrails, fallback logic, and system prompts ensure the chatbot remains grounded, professional, and strictly focused on career guidance.
Frontend:
The user interface is developed with Next.js (React) and styled using Tailwind CSS, providing a responsive, modern experience across devices. Key features include a real-time chat interface with session memory, new/reset chat controls, save/delete history options, and a frictionless “Try the Chatbot” mode that doesn’t require login. React hooks manage chat state and API interactions, while markdown rendering ensures clean, readable responses. The UI was purposefully designed with gradient backgrounds, accessible typography, and iconography to create an inviting experience.
AWS Deployment:
The entire application stack is hosted on an AWS EC2 instance running Ubuntu. Deployment is handled via a custom shell script (`post-checkout-setup.sh`), which installs dependencies and launches the frontend in a persistent `tmux` session. The backend FastAPI server is exposed on port 8000 and communicates with the frontend via REST APIs. DynamoDB is used to store and retrieve chat history, enabling persistent memory across user sessions. Logs are written to disk for monitoring and debugging, and the infrastructure is lightweight, scalable, and cost-effective.
Data Visualization:
To make labor data more accessible, Career Compass includes a rich Tableau-based Career Insights Dashboard. Raw occupational data from O*NET was supplemented with employment and wage data from the U.S. Bureau of Labor Statistics. After cleaning and merging the datasets using Standard Occupational Classification (SOC) codes, we created:
- An "Overview Dashboard" visualizing top job categories, average wages, and employment trends.
- Two narrative data stories:
- “Wages in America” showing wage distribution across roles and regions.
- “Employment Trends” highlighting job growth and decline over time.
Visualizations are embedded via responsive iframes and styled using Tailwind’s utility classes. Interactivity is enhanced with dropdown filters, tooltips, and clearly written insights to help users explore labor trends with ease. The visual experience is tailored for non-experts, enabling even first-time users to understand and navigate complex datasets.
Overall, Career Compass offers a powerful fusion of conversational AI, real-world job data, and interactive dashboards—distinguishing itself from traditional job sites by making career exploration intelligent, engaging, and data-driven.