Colin Taylor
Menlo Park, California, United States
Summary
🤩
RockstarColin Taylor is a Staff Software Engineer with 12 years of experience, currently building and shaping PyTorch at Meta from Menlo Park. He blends deep ML systems expertise with pragmatic engineering—contributing analytics logging, distributed training improvements, and API compatibility work to core PyTorch and TorchRec. His background includes founding a startup and technical leadership at Headspace, reflecting both product-minded engineering and operational experience. An MIT-trained AI and CS specialist, he focuses on readability, maintainability, and composability in high-performance GPU code, and has improved training pipelines and sharded embedding APIs that power large-scale recommendation and transformer workloads.
12 years of coding experience
Github Skills (11)
pytorch10
machine-learning10
distributed-training10
recommendation-system10
python10
analytics10
fs9
deep-learning9
tensor8
cuda8
autograd8
Programming languages (5)
Github contributions (5)
pytorch/torchrec
Nov 2021 - Jan 2023
Pytorch domain library for recommendation systems
Role in this project:
Back-end Developer
Back-end DeveloperContributions:3 releases, 47 reviews, 146 commits in 1 year 2 months
Contributions summary:Colin's commits primarily focused on improving the readability and maintainability of the `train_pipeline` module. They made code adjustments to clarify the device transfer, data distribution, and forward/backward stages within the training pipeline. Additionally, the user added license information to various files and corrected setup configurations. This involved modifying files related to optimization and distributed training aspects of the recommendation system library.
cudapytorchrecommendation-systemsdeep-learninggpu
pytorch/pytorch
Dec 2021 - Jan 2023
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:ML Engineer
ML EngineerContributions:26 reviews, 7 commits, 11 PRs in 1 year 1 month
Contributions summary:Colin contributed to the PyTorch library by adding analytics logging, specifically focusing on logging API usage for various modules and functionalities. They implemented analytics callsites for transformers, encoder/decoders, `torch.save` and `torch.load` functions, and the `torch._dynamo.optimize` and `torch._dynamo.export` functionalities. Furthermore, the user updated the `ShardedEmbeddingBagCollection` to be compatible with composable APIs and added a test for `ignored_sharded_tensor`. Additionally, they made the contract() function pickle-able for use with torch package, and integrated analytics logging to PT-D composable APIs.
pythongpu-accelerationdeep-learninggpunumpy
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.