Handwritten digit recognition requires robust models that work across varied image dimensions and real-world inputs

3-layer neural network with image normalization, grayscale conversion, resizing, inversion, and center of mass alignment. Trained using Adam optimizer achieving peak accuracy in 5 epochs

Improved classification accuracy from 92% to 98% on real-world inputs, achieving peak accuracy in less than 30 seconds

Nutrition tracking requires accurate calorie estimation across diverse food items and serving sizes

Full-stack application combining frontend, backend, SQL database, and USDA food database for comprehensive food search system

Led complete project lifecycle from ideation through deployment, achieving 85-95% accuracy for common foods

Generic workout plans don't adapt to individual needs, goals, and constraints for athletes

AI-powered recommendation system that generates personalized gym and cardio routines based on user preferences and goals

Created personalized routines for 20+ teammates during swim offseason, helping maintain strength and conditioning. Resulted in personal weight loss of over 10 lbs

Users need an efficient way to manage, track, and organize their reading lists and book collections

Full-stack application with intuitive interface for book discovery, tracking reading progress, and managing personal libraries

Provides users with a streamlined solution for organizing their reading journey and discovering new books

Wordle games lack intelligent computer opponents with varying difficulty levels and adaptive hint systems

Python-based Wordle implementation with traditional gameplay and AI computer mode featuring multiple guessing algorithms and advanced hint generation based on user input

Created an engaging word puzzle game with competitive AI opponents, demonstrating algorithm design and game development skills