✕ Close
Close

Contact

Feel free to reach out about research, projects, or anything you found intriguing.

Click to copy Copied.

Open in mail

LinkedIn ↗ · CV

Skip to content
TOOL mark

TOOL

An end-to-end music generation and style-transfer platform trained on a custom audio and lyrics scrape, with sweeps across 12+ model variants.

Contact about this work /contact

ML engineer

2024

Python, PyTorch, TensorFlow, TensorBoard

01The problem

I wanted generated music that kept rhythm and harmony instead of collapsing into texture, and I needed a pipeline I could actually train and compare.

02What I did

I built a custom audio and lyrics scraper for dataset ingestion, then trained U-Net-inspired generative models with mixed-precision FP16. TensorBoard tracked sweeps across 12+ architectural variants aimed at rhythmic and harmonic fidelity.

03How it works

Python, PyTorch, and TensorFlow. FP16 training to fit the sweeps. TensorBoard for comparing variants. The scraper and the training loop are part of the same platform, so a data change can be measured in the next run.

04What happened

TOOL is a working generation and style-transfer stack, not just a notebook. This page is a stub. I will write the full case study later.

Build real things. Make them worth using.

About Contact GitHub ↗ Scholar ↗ CV