Toby Boyd
Technical Program Manager (Gemini Data)
Ohio, United States
Summary
🤩
Rockstar🎓
Top SchoolToby Boyd is a pragmatic Technical Program Manager with nine years of experience driving high-impact ML infrastructure and performance programs at Google DeepMind and Google, and earlier product work at Amazon. He specializes in walltime and inference performance for large models—owning performance frameworks, ablation pipelines, and the data-selection process for Gemini pre-training—and has shipped improvements that materially reduced TPU costs and sped deployment across Bard, Search, and Assistant. Equally comfortable writing code and running operations, he has concrete open-source contributions to TensorFlow (models, agents, examples) focused on input-pipeline optimization, multi-GPU performance, and MKL/GPU support. Toby builds lightweight, outcome-focused teams and tooling that turn experimental ideas into repeatable, auditable wins, and he navigates complex legal and organizational constraints to get work into production. Based in Columbus, OH, he pairs an unusually hands-on engineering temperament with program-level leadership and a personal drive to push himself to the limit before stepping back.
9 years of coding experience
16 years of employment as a software developer
Bachelor Economics, Bachelor Economics at The College of Wooster
Github Skills (14)
machine-learning10
tensorflow10
python10
model-optimization10
reinforcement-learning10
testing9
performance-tuning9
gpu-programming9
dqn9
devops8
data-pipelines8
automated-build8
documentation8
data-pipeline8
Programming languages (6)
Github contributions (5)
tensorflow/agents
Nov 2019 - Jan 2023
TF-Agents: A reliable, scalable and easy to use TensorFlow library for Contextual Bandits and Reinforcement Learning.
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:220 commits, 46 PRs, 57 pushes in 3 years 2 months
Contributions summary:Toby's commits primarily focused on improving the usability and functionality of the `tensorflow/agents` repository. The user implemented a no-op method to resolve issues with network summaries and introduced benchmarking capabilities for DQN agents. Further, the user made crucial updates to the release process and Colab tutorials, indicating involvement in build, test, and documentation. This demonstrates contributions that span both back-end development and operational aspects.
scalabletf-agentsmultiagent-reinforcement-learningtensorflow-librarymulti-armed-bandits
tensorflow/models
Mar 2017 - Sep 2019
Models and examples built with TensorFlow
Role in this project:Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:183 commits, 331 PRs, 192 pushes in 2 years 6 months
Contributions summary:Toby's primary contributions focused on improving the input pipeline for CIFAR-10 by optimizing CPU performance and increasing throughput. They implemented changes to the CIFAR-10 model, specifically in the `cifar10_multi_gpu_train.py` and `cifar10_train.py` files. These changes included modifying the input pipeline to run on the CPU and prefetching data to improve performance. They also addressed and fixed typos. Finally, the user added MKL support and made adjustments to the code to work with GPUs.
deep-learningtensorflow
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.