James Betker
Research Engineer at OpenAI
Boulder, Colorado, United States
Summary
🤩
Rockstar🎓
Top SchoolJames Betker is a Systems Analyst with 14 years of experience based in Colorado who blends systems thinking with hands-on ML engineering. He contributes to open-source generative audio work, notably the multi-voice tortoise-tts project, where he focused on adapting a discrete diffusion vocoder, residual blocks, attention mechanisms, and spectrogram priors to improve audio quality. Self-described as a "Latent Analyst, Entropy Wrangler," he specializes in turning probabilistic, architecture-level ideas into robust model components that bridge research and production. His background suggests a pragmatic knack for shipping complex ML systems that sit at the intersection of signal processing and scalable system design.
14 years of coding experience
10 years of employment as a software developer
UC Santa BarbaraBS, Computer Science, BS, Computer Science at University of California, Santa Barbara
Stackoverflow
61reputation
2kreached
0answers
3questions
Github Skills (12)
machinelearning10
diffusion-models10
pytorch10
machine-learning10
audio-processing10
deep-learning10
python10
transformer-models6
huggingface-transformers6
keras6
nlp6
tensorflow6
Programming languages (5)
Github contributions (5)
neonbjb/tortoise-tts
Jan 2022 - Jan 2023
A multi-voice TTS system trained with an emphasis on quality
Role in this project:
ML Engineer
ML EngineerContributions:12 reviews, 168 commits, 32 PRs in 1 year
Contributions summary:James's commits focus on the development and modification of a discrete diffusion vocoder model within the tortoise-tts project, suggesting a focus on audio generation. Their work involves implementing and modifying residual blocks, attention mechanisms, and timestep embeddings, indicating expertise in the core architecture of the diffusion model. The commits demonstrate the user's direct involvement in adjusting the model to work with a spectrogram prior, which helps with improving audio quality.
pytorchvoicequalityemphasistts
neonbjb/ocotillo
Nov 2021 - May 2022
Performant and accurate speech recognition built on Pytorch
Contributions:3 releases, 48 commits, 25 pushes in 6 months
pytorchspeech-to-textrecognitionctcspeech-recognition
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.