Nat Jeffries
Machine Learning Researcher at Useful Sensors
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolNat Jeffries is a Machine Learning Engineer with 11 years of experience building production-ready ML systems, from TinyML kernels at Google to insulin dosing algorithms at Replica Health. Based in Mountain View, he has deep embedded ML expertise—contributing an optimized int8 conv kernel for the widely used tflite-micro project and helping drive TinyMLPerf benchmarking—while also applying large-scale data and simulation to physiological modeling for Type 1 Diabetes. A pragmatic engineer and occasional academic collaborator (visiting scholar at Stanford), Nat combines low-level performance tuning with end-to-end evaluation pipelines that bridge virtual patients and real-world verification. He previously co-founded Useful Sensors and has a strong background in electrical and computer engineering from Carnegie Mellon, bringing both startup grit and production discipline to challenging ML deployment problems.
11 years of coding experience
9 years of employment as a software developer
Master of Science (MS), Electrical and Computer Engineering, Master of Science (MS), Electrical and Computer Engineering at Carnegie Mellon University's College of EngineeringAmherst Regional High School
Chinese
Github Skills (10)
neural-network10
embedded10
machine-learning10
convolutional-neural-networks10
tensorflow-lite10
sys10
xtensa10
cprogramming-language9
microcontroller9
c-language9
Programming languages (4)
Github contributions (5)
tensorflow/tflite-micro
Jun 2021 - Aug 2022
Infrastructure to enable deployment of ML models to low-power resource-constrained embedded targets (including microcontrollers and digital signal processors).
Role in this project:
ML Engineer
ML EngineerContributions:236 reviews, 31 commits, 116 PRs in 1 year 2 months
Contributions summary:Nat primarily contributed to the implementation of an int8 conv kernel variant for the Xtensa architecture within the tflite-micro framework. They optimized binary size and addressed comments, renaming and refactoring code. Their work also involved integrating the new kernel within the person detection benchmark and creating supporting registration functions.
signalml-modelslow-powerprocessorsdeployment
njeffrie/tiny
Jun 2020 - Aug 2021
Contributions:2 PRs, 37 pushes, 5 branches in 1 year 1 month
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.