An ML-powered nutrition assistant that turns photos of food into nutrition estimates, helping users make more informed dietary decisions.
Spearheaded dataset acquisition and preprocessing for image-based nutrition analysis, integrating scraping, labeling, augmentation, and quality checks to improve training signal and data reliability.
Built an end-to-end machine learning pipeline for image classification using transfer learning with systematic experiment tracking to converge on the best-performing model.
Developed a standardized image preprocessing workflow covering cleaning, cropping, resizing, normalization, and augmentation to improve model robustness and generalization.
Implemented HDF5 model export and served inference through a lightweight Flask API, enabling Cloud and Android integration.