PathFinder AI: A Three-Stage Machine Learning Pipeline for Predicting Graduates' Employment and Salary

PathFinder AI is a full-stack predictive analysis system for college graduates, addressing employment market information asymmetry. It integrates academic data, internship experience, and skill assessments to provide personalized career path planning and salary expectations. The system uses a three-stage ML pipeline (random forest, multiple linear regression, KNN) and is built with React/Vite/TypeScript frontend and Flask backend with scikit-learn models.

Graduates face two main challenges: difficulty in self-positioning in the job market and lack of objective salary references. Traditional career consulting relies on experience and can't quantify individual differences. PathFinder AI aims to solve these via data-driven personalized advice. The system's architecture ensures real-time predictions and scalability (React frontend, Flask backend).

The core of PathFinder AI is its three-stage pipeline: 1. Employment Probability Classification: Random Forest classifier uses GWA, internship parameters (duration, company size, role match), and academic backlog to assess employment probability (captures non-linear feature interactions). 2. Salary Regression Prediction: Multiple linear regression adjusts entry-level salary based on company size, industry, location, and job type (linear relationships and explainable coefficients). 3. Career Path Optimization: KNN algorithm recommends optimal paths based on skill alignment (Python, DSA, Web dev, ML) by comparing user skills with successful cases (similarity-based, interpretable).

Frontend uses React/Vite/TypeScript for type safety and fast iteration. Backend Flask API follows RESTful principles; models are serialized with joblib for efficient loading. Data preprocessing handles heterogeneous data (academic from教务, internships from就业 center, skills from tests/portfolios) via cleaning, feature engineering, and standardization to ensure model performance.

PathFinder AI serves multiple stakeholders: 1. Graduates: Personalized career and salary guidance. 2. University Career Centers: Identify students needing extra support. 3. Enterprises: Optimize campus recruitment strategies. 4. Education Institutions: Data to adjust curricula based on market demand gaps (via prediction vs actual employment deviations).

Current limitations: Data privacy concerns (sensitive student data requires strict access control/encryption) and model timeliness (needs regular retraining with latest market data). Future improvements: Integrate deep learning for complex feature interactions, NLP for resume analysis, and real-time feedback loops for model refinement.

PathFinder AI demonstrates how classic ML algorithms (random forest, linear regression, KNN) can form a complete system to solve real-world education/employment problems. Its three-stage design and full-stack implementation provide a valuable reference for ML engineering practitioners, showing the path from algorithm prototype to usable product.