Daniel Herrington

I am an engineer who builds and scales machine learning systems that make life simpler. Currently Chief AI Officer @ The Zebra, previously a product manager on the Gemini infrastructure team at Google.

Daniel Herrington Profile Headshot

I am a software engineer and product leader focused on building and scaling large machine learning systems. Over my career, I've worked across startup trenches and massive cloud infrastructures, bridging the gap between deep technical systems and intuitive customer products.

2026 - Present
The Zebra Logo
I am currently the Chief AI Officer (CAIO) at The Zebra, the nation's leading insurance comparison marketplace. Here, I lead artificial intelligence product strategy, data science, and engineering to simplify how millions of users compare and buy insurance. My day-to-day focus centers on fine-tuning models, operating custom AI agent architectures to level-up our customer experience, and building scale-ready, enterprise-wide AI governance frameworks.
Before 2026
Google Logo
I led key product efforts in Google's AI infrastructure division, building the systems that power modern warehouse-scale machine learning. As a product leader for the CoreML frameworks team and critical serving infrastructure for Google's next-generation Gemini models, I partnered with researchers across Google DeepMind and Google Research to enable robust, ultra-fast model training pipelines. During this time, I authored the ML Productivity Goodput framework and led the development of its open-source measurement library in the google/AI-Hypercomputer repository to optimize production TPU/GPU fleet efficiency.
Prior Experience
Vroom Logo Capital One Logo Priceline Logo
Before Google, I spent over a decade designing, operating, and scaling complex technical product architectures. At Vroom, Capital One, and Priceline, I drove advanced digital products, machine learning initiatives, massive A/B testing engines, and platform scaling. During this journey, I co-founded two startups, successfully raised venture capital, and led a consumer product launch that attracted international media attention, peaked at #3 in the App Store, and served over 500,000 active customers on its opening weekend.
Education
Mississippi State University Logo
I completed my B.S. in Electrical Engineering at Mississippi State University, focusing my academic research on signals and systems, control systems, and computational frameworks. This work built the foundational mathematics and analytical principles that still guide my approach to deep learning systems and computing clusters today.

Writings & Insights

Research & Systems Engineering

Machine Learning Fleet Efficiency Abstract Visualization
Research Paper

Machine Learning Fleet Efficiency: Analyzing and Optimizing Large-Scale Google TPU Systems with ML Productivity Goodput

Introduces a systematic framework (MPG) to measure and optimize the efficiency of warehouse-scale ML infrastructures, analyzing system-wide dependencies across Google's production TPU fleet.

Read on arXiv
ML Productivity Goodput Metrics Abstract Visual
Google Cloud Blog

Goodput metric as measure of ML productivity

Introducing the ML Productivity Goodput framework on Google Cloud, decomposing training efficiency into Scheduling, Runtime, and Program metrics to pinpoint and eliminate compute waste.

Read on Google Cloud Blog
ML Goodput Python Library Abstract Visual
Open Source

ML Goodput Measurement Library

A Python-based measurement library inside Google's AI-Hypercomputer repository to instrument ML models and automatically log productive training time and overhead.

View on GitHub
daniel@herrington:~/dh.sh
StdioTransport Active
developer-logs.sh
dh.sh Stdio Proxy