David Holtz
Machine Learning Ops Engineer at Hugging Face
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolDavid Holtz is a Machine Learning Ops Engineer based in New York with 11 years of experience building high-performance, production-ready systems across Rust, CUDA, Go, TypeScript, and Python. Currently a core maintainer on Hugging Face’s Text Generation Inference, he enabled OpenAI-compatible chat completions and features that make TGI a drop-in replacement for OpenAI clients, reflecting his focus on real-world ML deployments and multi-GPU optimization. His background spans blockchain and crypto infrastructure—designing a Substrate-based DEX and high-throughput data pipelines at Chainlink—bringing strong security and systems design to ML ops. David combines low-level ML engineering (contributions to a Rust minimalist ML framework implementing VGG models and backprop for custom activations) with practical product delivery from wallets to serverless education services. He’s equally at home writing CUDA kernels or clear technical prose, and his GitHub emoji-laced bio hints at a pragmatic, curious engineer who enjoys working at the intersection of systems, ML, and tooling.
11 years of coding experience
9 years of employment as a software developer
Master’s Degree Business Analytics (MS-BA), Master’s Degree Business Analytics (MS-BA) at W. P. Carey School of Business – Arizona State University
Spanish, English
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (23)
pytorch10
back-end-development10
apidoc10
http10
machine-learning10
inference10
jinjava10
implement10
deep-learning10
api10
backpropagation10
computer-vision10
rust10
modeling10
data-structure9
Programming languages (19)
Github contributions (5)
huggingface/text-generation-inference
Aug 2023 - Dec 2025
Large Language Model Text Generation Inference
Role in this project:
Back-end Developer
Back-end DeveloperContributions:2 releases, 383 reviews, 328 PRs in 2 years 4 months
Contributions summary:David implemented the OpenAI chat completions API support, enabling the Text Generation Inference (TGI) to be a drop-in replacement for OpenAI clients. This involved integrating the tokenizer configuration, applying chat templates, and adapting existing generate functions to align with the expected HTTP interface. Furthermore, the user introduced support for custom Jinja functions, including `raise_exception`, and incorporated BOS and EOS tokens into the template. The user's work facilitates compatibility and streamlines integration with OpenAI-based tools and frameworks.
inferencelarge-language-modelstext-generationbloomnlp
huggingface/candle
Aug 2023 - Jan 2026
Minimalist ML framework for Rust
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:9 reviews, 12 PRs, 2 pushes in 2 years 5 months
Contributions summary:David primarily contributed to the implementation of VGG models (VGG13, VGG16, and VGG19) within the candle framework. This involved defining the model architecture, including convolutional blocks and fully connected layers, using the provided codebase. The user added backpropagation support for functions, like erf and gelu-erf, expanding the capabilities of the framework. Furthermore, they enhanced the project by downloading and integrating CIFAR dataset parquet files and implementing enhancements and fixes to the audio processing pipeline.
machine-learningrust
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.