FROM RECRUITMENT DATA SOURCES TO PLACEMENT INTELLIGENCE — AN AUTOMATED ETL PIPELINE FOR CAMPUS HIRING ANALYTICS AND DECISION SUPPORT - (HIRESTREAM)

Abstract

Campus placement is one of the most closely watched activities in an engineering college, yet the data behind it is usually scattered across many places. Student academic records sit in the college ERP, resumes arrive as PDF files, company job descriptions come by email, aptitude test scores are exported from online assessment platforms, and interview outcomes are noted in spreadsheets by the placement cell. Preparing a simple report on how many students are eligible for a drive or why a department's placement rate fell requires days of manual copying. This paper presents HireStream, an automated extract, transform, and load pipeline that brings these recruitment data sources together and turns them into placement intelligence for the training and placement office. The pipeline extracts data on a schedule from the ERP database, a shared folder of resumes, a job description inbox, assessment platform exports, and interview feedback forms. Orchestrated by Apache Airflow, the transform stage standardises roll numbers, branch codes, and dates, validates records against rules for completeness and range, parses resumes with a spaCy-based information extraction model, and maps free-text skills onto a common skill taxonomy. The cleaned data is loaded into a PostgreSQL warehouse designed as a star schema, with fact tables for applications, assessment attempts, interview rounds, and offers, and dimension tables for students, companies, drives, skills, and time. Several analytical models run on the warehouse. A gradient boosting classifier estimates each student's probability of receiving an offer from features such as CGPA, backlogs, aptitude percentile, coding test score, internship experience, and matched skills. A matching model compares student skill profiles with job descriptions using TF-IDF and sentence embeddings and ranks eligible students for each drive. A skill gap analysis compares the skills demanded by recruiters over recent seasons with those present in each batch, and a funnel analysis shows where candidates drop out between registration, test, interview, and offer.