Developed "Humming AI," a music-production service where AI generates accompaniment from a recorded hummed melody, then lets you refine it interactively into your ideal song — built with Team Flex (3 people). Won an award.
A service that captures a melody just by recording you humming it, then evolves the song through natural conversation with AI — "make it more rock," "add drums" — resulting in an original track that keeps your own melody. Existing music-generation AIs (Suno, Udio, and others) can't preserve the original melody or support interactive editing; keeping the hummed melody at the center of the product was our point of differentiation.
Joined the 3-person team from the ideation stage and handled implementation. Worked on building the AI-driven architecture, including a Melody Service that denoises the hummed audio and estimates pitch, and a Backend for Frontend that integrates with the accompaniment-generation AI model (MusicGen).