Geoffrey Angus
Technical Lead, Machine Learning at Predibase
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolGeoffrey Angus is a Technical Lead in Machine Learning based in San Francisco with eight years of experience building and scaling ML systems from research to production. A Stanford MS/BS graduate, he has driven major performance and capability gains at Predibase—launching the LoRAX Multi-LoRA serving framework, overhauling the fine-tuning engine for 6x throughput, and leading initiatives that show small task-specific models can outcompete GPT-4 on niche tasks. His background includes increasing training throughput and cutting memory use for Transformer-based Image Search at Google and improving clinical mammogram screening accuracy with unsupervised approaches in healthcare. An active open-source contributor, he’s enhanced Ludwig’s text processing pipeline (TorchScript SentencePiece, Hugging Face tokenizer support, checkpoint loading) to make custom LLM workflows more robust. Fluent in both research and product delivery, he combines low-level systems optimizations with practical ML platform design, and has a proven knack for translating academic ideas into production wins.
8 years of coding experience
7 years of employment as a software developer
High School, High School at Heritage High SchoolMaster of Science - MS Computer Science, Master of Science - MS Computer Science at Stanford University
English, French, Tagalog, Spanish, Portuguese
Github Skills (10)
transformers10
torchscript10
huggingface-transformers10
pytorch10
machine-learning10
nlp10
python10
natural-language-processing10
text-processing10
tensorflow7
Programming languages (4)
Github contributions (5)
ludwig-ai/ludwig
Mar 2022 - Jan 2023
Low-code framework for building custom LLMs, neural networks, and other AI models
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:1 release, 271 reviews, 524 commits in 10 months
Contributions summary:Geoffrey's commits primarily focused on enhancing text processing capabilities within the Ludwig framework. The contributions involve the implementation of a Torchscript-enabled SentencePiece Tokenizer, the correction of prediction postprocessing within the text feature, and the integration of user-defined Hugging Face BERT tokenizers. These modifications improve the functionality and performance of the framework's text-based features for custom LLMs and other AI models. In addition, the user worked on creating the ability to load models using a training checkpoint.
fairness-mlpythonframework-learningdeep-learning-frameworknatural-language-processing
geoffreyangus/cs106r
Jun 2018 - Oct 2020
An introduction to Computer Science taught in Python.
Contributions:306 pushes, 1 branch, 3 issues in 2 years 4 months
introduction-to-computer-sciencepythoncomputer-sciencescience
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.