Rafael Valle
Chief Science Officer at Rime
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolRafael Valle is a research-driven machine learning scientist and engineering leader with 13 years of experience bridging audio, music, and multimodal generative models, now working at Meta Superintelligence Labs after an eight-year research leadership stint at NVIDIA. He holds an interdisciplinary PhD from UC Berkeley in machine listening and improvisation and brings deep domain expertise in music signal processing, MIDI handling, and text-to-speech systems—contributions include meaningful fixes and inference improvements to NVIDIA’s Flowtron and enhancements to the widely used pretty-midi library. Rafael combines rigorous academic training with production-minded research, shipping robust model and data-handling improvements that address precision, padding, and attention-related issues. Based in the San Francisco Bay Area, he pairs a rare background in orchestral conducting and computer music with applied ML, enabling creative approaches to multimodal generation and machine improvisation.
13 years of coding experience
12 years of employment as a software developer
Interdisciplinary PhD Machine Listening and Improvisation, Designated Emph. in Computational, Data Science and Engineering, Interdisciplinary PhD Machine Listening and Improvisation, Designated Emph. in Computational, Data Science and Engineering at University of California, BerkeleyMaster Computer Music / Composition, Master Computer Music / Composition at MH-StuttgartBachelor's in Orchestral Conducting Music Performance, General, Bachelor's in Orchestral Conducting Music Performance, General at Universidade Federal do Rio de Janeiro
Portuguese, Spanish, German, Italian, French, Hindi
Stackoverflow
6,883reputation
4.2mreached
52answers
7questions
tensorflow
top-5%
neural-network
top-5%
matplotlib
top-1%
data-visualization
top-1%
deep-learning
top-5%
python
top-5%
Github Skills (31)
algorithms10
pytorch10
python10
matplotlib10
attention-mechanism10
machine-learning10
speech-synthesis10
midi10
midi-instrument10
data-visualization10
data-structure9
pandas9
numpy9
data-modeling9
model-optimization9
Programming languages (11)
Github contributions (5)
NVIDIA/flowtron
May 2020 - Nov 2022
Flowtron is an auto-regressive flow-based generative network for text to speech synthesis with control over speech variation and style transfer
Role in this project:
ML Engineer
ML EngineerContributions:28 commits, 6 PRs, 40 pushes in 2 years 6 months
Contributions summary:Rafael primarily contributed to the `flowtron` repository by modifying the core files, `flowtron.py` and `data.py`. They focused on addressing potential issues related to data handling, padding, and ensuring compatibility with floating-point precision, especially concerning the attention mechanisms. The user also improved the inference capabilities with changes made to `inference.py`, including adding the `torch.no_grad()` context manager for waveglow inference.
speechstyle-transfertext-to-speechspeech-synthesis
craffel/pretty-midi
Jul 2015 - Dec 2015
Utility functions for handling MIDI data in a nice/intuitive way.
Role in this project:Back-end Developer
Back-end DeveloperContributions:7 commits, 13 PRs, 47 comments in 4 months
Contributions summary:Rafael primarily contributed to the core functionality of the `pretty-midi` library, focusing on features related to MIDI data handling. Their work included adding and refining features for retaining key and time signatures, crucial for accurate MIDI file processing. The user also added a utility to convert quarter notes per minute to beats per minute and implemented functionality for pitch class analysis. They demonstrated a strong understanding of MIDI data structures and algorithms.
midi
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.