Multi-Modal Data Extraction and Analysis of Digital Labour-Market Intermediaries

Abstract

Digital platforms increasingly mediate access to employment opportunities, yet their role in shaping labor market inequalities remains insufficiently understood. This thesis presents a comprehensive analysis of Digital Labor Market Intermediaries (DLMIs) across France and Brazil, focusing on care work platforms where gendered and racialized inequalities are particularly pronounced. Through innovative methodological approaches combining large-scale web scraping, multimodal AI inference, and explainable machine learning techniques, the research reveals systematic patterns of algorithmic bias in worker representation, pricing mechanisms, and platform success factors. This internship developed a scalable data engineering and AI pipeline to study how digital intermediation platforms shape labor market inequalities. By reverse engineering diverse platform APIs, the system automated extraction and quality validation of tens of thousands of caregiver proőles across multiple countries. Integrated multimodal AI models inferred rich demographic attributes, enabling large-scale, explainable analysis of algorithmic bias in pricing, temporal advantages, and geographic concentration. The project creates a foundation for real-time monitoring, fairness-aware recommendation systems, and evidence-based policy guidance to promote accountability and equity in the digital economy