Scam Detection Labelling
The project focused on building a high-quality labeled dataset for training and evaluating machine learning models to detect fake or inauthentic social media profiles. The work involved annotating and validating structured profile data using both manual review and AI-assisted classification. Labeling Tasks • Classified accounts into defined categories (e.g., authentic vs fake) • Applied structured tags and metadata to support model training • Reviewed edge cases and resolved ambiguous classifications • Performed cross-model validation using multiple LLM outputs Project Size & Workflow • Processed large CSV-based datasets containing thousands of records • Implemented batch labeling and automated pipelines to improve efficiency • Merged and cleaned datasets to ensure consistency before annotation • Generated structured outputs ready for machine learning ingestion