CareerPath AI
Released: Nov 2025
Description
CareerPath AI serves as a proof-of-concept for the potential of 'vibe coding' with next-gen models. Built entirely over a single three-day weekend using Gemini 3 Pro, this platform acts as an intelligent career counselor that moves beyond generic advice. It ingests user demographics and quiz results to generate hyper-personalized career roadmaps. The system helps users visualize their future by generating realistic 'day-in-the-life' image slideshows using AI, bridging the gap between abstract job titles and daily reality. This project demonstrates how modern AI tools can accelerate the development of complex, full-stack applications with robust architectures.
Key Responsibilities
Prompt-engineered the entire full-stack codebase using Gemini 3 Pro, acting as the architect while the AI handled the implementation details.
Deployed the backend to Hugging Face Spaces using Docker and the frontend to Vercel, ensuring a seamless global delivery pipeline.
Designed the comprehensive database schema and Row Level Security policies to ensure strict user data privacy within Supabase.
Orchestrated the asynchronous image generation pipeline, coordinating multiple parallel API calls to reduce wait times for the end user.
Implemented a robust state management system using Zustand to handle the complex user journey from onboarding to result visualization.
Technology Used
React
TypeScript
TailwindCSS
Python
Supabase
PostgreSQL
Google Gemini
Docker
Features
Deep Profile Analysis
The core engine utilizes Google Gemini 2.5 Flash to reason through user data, creating matches based on aptitude and personality rather than simple keyword associations.
Actionable Roadmaps
Users receive step-by-step educational and professional plans detailing exactly how to transition from their current state to their target career, including location-specific advice.
Generative Day-in-the-Life
A unique visualization feature (powered by Google's Nano Banana) that generates realistic, non-glamorized photo slideshows of what a specific job looks like daily, helping users emotionally connect with potential career paths.
Decoupled Architecture
The application uses a secure "headless" structure with a React frontend and a Python FastAPI backend acting as a proxy to manage API keys and business logic securely.
Smart Rate Limiting
To maintain sustainability, a custom database-level quota system manages daily limits for expensive operations like image generation and detailed career analysis.
Screenshots
Additional Information
This project was an experiment in "vibe coding" to test the reasoning capabilities of Gemini 3 Pro. The entire application was built in just 3 days with the AI generating nearly all the code.
The "intelligence" of the app relies on rigid JSON-enforced prompt engineering, which forces the LLM to output structured data that the frontend can reliably render.
Unlike simple API wrappers, this project implements a full backend proxy to secure the Gemini API keys, preventing them from being exposed in the client-side code.
I implemented a "lazy-loading" strategy for the career details to save on token costs; detailed roadmaps are only generated when a user explicitly clicks to explore a specific career.
The UI features a "Nano Banana" badge in the loading screens, a playful nod to the internal codename and the rapid, fluid nature of the development session.