• Built and deployed machine learning models to recommend recipes using structured data (user preferences, dietary tags, ingredient vectors), improving personalization and engagement.
• Designed a content-based filtering system using cosine similarity and feature encodings (TF-IDF, one-hot, scaled nutrition scores) to generate top-N recipe suggestions.
• Developed modular training pipelines in scikit-learn and automated preprocessing workflows for feature generation, normalization, and input validation.
• Implemented offline evaluation pipelines using metrics such as precision, recall, and coverage, and conducted A/B testing to benchmark different filtering strategies.
• Created interactive dashboards using Power BI to track model performance, ingredient frequency, and user interaction patterns, enabling real- time feedback for iterative model tuning.
• Ensured data quality by implementing validation checks and schema consistency rules during data ingestion, increasing model stability and trustworthiness.
• Focused on reproducibility and scalability by maintaining clean code, consistent documentation, and experiment tracking for all modeling iterations.