Optimizing Machine Learning Models for Gen AI Workflows

This technical report outlines the implementation of a machine learning optimization project by Vaco for a multinational IT software and services company. The project achieved a quality rating of 99.5% within six weeks, focusing on enhancing AI-generated advertising through a robust training model. The initial phase concentrated on English solutions, leading to an expansion into 14 additional languages. Vaco established an incubation workflow involving 60 onshore human reviewers to ensure high-quality outputs and develop foundational policies aimed at risk reduction and brand protection. The report details the guidelines created, the analysis of AI-curated content, and the establishment of a policy feedback loop with internal stakeholders to manage changes effectively. Key priorities included maintaining user experience quality and retaining third-party advertisers while addressing challenges related to cultural relevance and compliance with geographical regulations. The document emphasizes the importance of human review in improving AI and machine learning models.