Maverick
Audio AI ยท DSP ยท ML
Bio
I'm a CS undergrad building a specialization in Audio AI โ the intersection of digital signal processing, machine learning, and sound. I'm interested in how machines hear: spectral analysis, speech enhancement, source separation, and audio classification.
Currently working through a self-directed curriculum that goes from raw NumPy DSP implementations (FFT, STFT, Mel filterbanks) up through neural audio models. I like understanding things from first principles before reaching for libraries.
Projects
- Mel Spectrogram Pipeline โ A from-scratch NumPy implementation of the full Mel spectrogram pipeline (windowing โ FFT โ STFT โ Mel filterbank โ log scaling), verified against Librosa. Built to deeply understand every step rather than call an API.
- Tylgo โ Production web platform built for a tiles showroom. Full-stack (Next.js, Supabase) with Tally ERP integration for automated stock and sales sync, canvas-based room dimension drawing, and interactive 3D room/staircase visualizers. Freelance project, actively deployed.
- Blindsight โ Chrome extension that reads and classifies Terms of Service documents using NLP, surfacing the clauses that actually matter. Built at Makeathon8 (MLSC).
Blog
I write on Blogspot.