Software

Most of the software projects I've worked on can be found in GitHub.

• GTN A Python and C++ library for automatic differentiation with weighted finite-state transducers ( documentation, blog post).
• Flashlight A lightweight and efficient library for machine learning written in C++.
• wav2letter++ A toolkit for state-of-the-art end-to-end speech recognition in C++.
• CrypTen A library for secure and privacy preserving machine learning built on top of PyTorch.
• RNN Transducer A fast C implementation of the RNN Transducer loss function with PyTorch bindings.
• ECGs Software to reproduce our work "Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network" (project webpage).
• Keyword Spotting A library for joint, end-to-end keyword spotting and voice activity detection in TensorFlow and the corresponding publication.
• warp-ctc A fast CUDA GPU and CPU implementation of the CTC loss function with a fork for PyTorch bindings. Details of the CUDA version can be found in the Appendix of Deep Speech 2.