Transformer from Scratch (In Progress)
Currently planning and designing an educational project to implement a Transformer model entirely from scratch in PyTorch, to deepen understanding of modern deep learning architectures.
Projects
My work spans from production-grade full-stack systems to experimental prototypes exploring ML and its applications.
Modal – AI Music Theory Discovery Platform
An AI learning platform that utilizes the Spotify 12M track dataset and Ultimate Guitar chord data (via scraping) to provide deep musical insight. Combines large-scale music data analysis with AI to guide learners through music theory concepts, chord progressions, and genre patterns. Not currently hosted due to licensing concerns.
Accent Learner – AI Pronunciation Coach
An AI pronunciation trainer that listens to your speech, compares it against native pronunciation models, and provides personalized coaching feedback. Uses advanced speech recognition to analyze phoneme accuracy, prosody, and language-specific patterns across five languages (English, French, Spanish, German, Italian). Combines automated speech recognition with intelligent tutoring to give learners the detailed phonetic feedback typically only available from human instructors.
From-Scratch Neural Network for MNIST
Implemented logistic regression, linear regression, and a feedforward neural network with backpropagation from scratch to train on the MNIST handwritten digit dataset, achieving strong accuracy while reinforcing deep understanding of machine learning fundamentals.