Nate Sutton
Nate Sutton
teaching machines about healthcare
Embedded Files
I'm a pragmatic machine learning engineer with 9 years of experience delivering data products into production. I enjoy working at the intersection of machine learning and product management to make an impact in front-line patient care. Currently I am building a startup from the ground up as the founding engineer at Inflo Health.
I live in the mountains of North Carolina and spend my off-hours shepherding 🐑.
EXPERIENCE
EXPERIENCE
2020 - present
Inflo Health identifies recommendations for additional imaging in radiology reports to ensure patients never miss a follow-up exam. In this role I brought a 0 to 1 product to market while leading a small development team.
Inflo Health identifies recommendations for additional imaging in radiology reports to ensure patients never miss a follow-up exam. In this role I brought a 0 to 1 product to market while leading a small development team.
2018 - 2023
Mentored 30+ students during their transition to data roles.
Mentored 30+ students during their transition to data roles.
2019 - 2020
Transformed our database pipelines to continuous integration and deployment, and this allowed our team to release features daily instead of bi-weekly.
Transformed our database pipelines to continuous integration and deployment, and this allowed our team to release features daily instead of bi-weekly.
2017 - 2019
Patented a novel caching strategy to identify patients with acute coronary syndrome in real-time (#11,177,041 ↗)
Patented a novel caching strategy to identify patients with acute coronary syndrome in real-time (#11,177,041 ↗)
2015 - 2017
Delivered 6 different acute-care models. This diversified our product from a single use case and led to our series A round.
Delivered 6 different acute-care models. This diversified our product from a single use case and led to our series A round.
PROJECTS
PROJECTS
2023
Over 10 papers in the ML literature in the last 5 years claim to generate accurate radiology reports from medical images, but they all suffer from the same flaw. None of them answer a clinical question. In this project I trained a BLIP model (specifically, a dual text/vision encoder paired with a text decoder) to generate more clinically useful radiology reports given a chest x-ray and a patient’s reason for visit.
Over 10 papers in the ML literature in the last 5 years claim to generate accurate radiology reports from medical images, but they all suffer from the same flaw. None of them answer a clinical question. In this project I trained a BLIP model (specifically, a dual text/vision encoder paired with a text decoder) to generate more clinically useful radiology reports given a chest x-ray and a patient’s reason for visit.
2022
When you have a relational database, you will inevitably have to change it! This repo will help you decide between 3 useful frameworks for database migrations by walking through each implementation in a dockerized PostgreSQL database.
When you have a relational database, you will inevitably have to change it! This repo will help you decide between 3 useful frameworks for database migrations by walking through each implementation in a dockerized PostgreSQL database.
This repository allows you to quickly spin up a set of common services that are useful in building machine learning proof of concepts. Several of my mentees have had success using this a launching point for take home assignments they were given during their application process.
This repository allows you to quickly spin up a set of common services that are useful in building machine learning proof of concepts. Several of my mentees have had success using this a launching point for take home assignments they were given during their application process.
2021
Thanks to new regulations, hospitals are required to disclose their prices for common services on their websites. I curated a list with every hospital in North Carolina and made their prices comparable in an open database.
Thanks to new regulations, hospitals are required to disclose their prices for common services on their websites. I curated a list with every hospital in North Carolina and made their prices comparable in an open database.
2018
Chest pain is one of the most common reasons for a patient to visit the emergency room, but few have a major cardiac event. At Vituity we patented a method for identifying these patients in real-time using a novel caching strategy. This supports clinicians in their decision to discharge low-risk patients from the ER and helps reduce overcrowding in admissions to higher levels of care.
Chest pain is one of the most common reasons for a patient to visit the emergency room, but few have a major cardiac event. At Vituity we patented a method for identifying these patients in real-time using a novel caching strategy. This supports clinicians in their decision to discharge low-risk patients from the ER and helps reduce overcrowding in admissions to higher levels of care.
I led Vituity's finalist submission for the Department of Homeland Security's challenge and was awarded $20K. Our solution monitored emergency department wait times across 142 hospitals in 19 states to detect emergent influenza pandemics.
I led Vituity's finalist submission for the Department of Homeland Security's challenge and was awarded $20K. Our solution monitored emergency department wait times across 142 hospitals in 19 states to detect emergent influenza pandemics.
SKILLS
SKILLS
Languages
SQL
Python
R
Platforms
Azure
AWS
Tools
Bash
Git
Docker
ML Frameworks
Pytorch
H2O
Domains
Machine Learning
Language
Time Series
Standards
HL7
FHIR
PUBLICATIONS
PUBLICATIONS
2018
Is conservation right to go big? Protected area size and conservation return-on-investment
2017
Factoring economic costs into conservation planning may not improve agreement over priorities for protection
2016
The grain of spatially referenced economic cost and biodiversity data and the effectiveness of a cost targeting strategy
2014
A reliance on agricultural land values in conservation planning alters the spatial distribution of priorities and overestimates the acquisition costs of protected areas
MENTEES
MENTEES
Where are they now?
Where are they now?
Page updated
Google Sites
Report abuse