Hello there! Welcome to a little landing page with some links to stuff I'm tinkering with.
Used to play guitar when I was young. Became interested in using machine learning for music understanding in computer science school around 2011 from the artificial neural networks side.
Picked up Theano when Lasagne/Keras was coming up and was a regular contributor to Keras and TensorFlow before 1.0. Sad about 2.0 as an aspiring functional programmer, but now moving to PyTorch like everyone else.
I've touched a lot of deep learning models, particularly around audio (most recently time-domain diffusion). My little claim to fame is polyphonic pitch detection expressed as semantic segmentation of synthesised piano rolls.
Generally interested in representation learning by self-supervision, and applying differentiable programming to digital signal processing for next-level audio products and music production tools.
I'm currently thinking about audio-based finetuning methods, reproducible code for scientific computing, and whether I should study philosophy & plumbing instead.
Look for carlthome at GitHub, Gmail, Reddit, Twitter, LinkedIn, Kaggle, Spotify, Facebook, SoundCloud, Bandcamp, last.fm, YouTube, etc. It's usually me. Sometimes it's another guy.
Some interesting workplaces I've been developing at:
- Epidemic Sound - 2023-now
- XLN Audio - 2022
- Epidemic Sound - 2019-2021
- Peltarion 2018-2019
- DoReMIR - 2016-2018
- [break for engineering school]
- Civil Rights Defenders 2010-2011
- DevCore - 2007-2008
Some cool people I've been fortunate to support on their learning journey:
- Sebastian Löf & Cody Hesse - Self-supervised learning of musical representations using VICReg
- Oriol Colomé Font - Uncovering underlying high-level musical content in the time domain
- Nicolas Jonason - The control-synthesis approach for making expressive and controllable neural music synthesizers
- Laura Cros Vila - Musical Instrument Recognition using the Scattering Transform
- Carl Nyströmer - Musical Instrument Activity Detection using Self-Supervised Learning and Domain Adaptation
- Emma Weberyd - Evaluating the Impact of Automated Music Tags on Search Engine Ranking Quality
- Waseem Sheriff - Learning to predict text quality using Generative Adversarial Networks
- Amund Vedal - Unsupervised Audio Spectrogram Compression using Vector Quantized Autoencoders