Kevin (KJ) Krause

Machine Learning Engineer

Summary

Machine Learning Engineer with a focus on biomedical informatics. Extensive research and application experience in ML, NLP, and sentiment analysis for healthcare.

Experience

[+0000bjeca Objeca] - [+0000bjeca Objeca]

Research Assistant

Developed improved ML tools focused on histology image analysis

[+0000bjeca Objeca] - [+0000bjeca Objeca]

Intern

Implemented an online data-entry process for QC’s equipment qualification (EQ). Collaborated with stakeholders to revise SOP’s in the interest of FDA compliance. Analyzed inventory data to detect and meliorate non-compliant instruments. Participated in LEAN operations and collected KPI’s to track and reduce waste.

[+0000bjeca Objeca] - [+0000bjeca Objeca]

Machine Learning Engineer / App Developer

Used Python to build powerful machine learning models for clients. Deployed applications to the web using attractive front-end interfaces. Designed and implemented back-end SQL databases and REST APIs. Emphasis on natural language processing and recurrent neural networks.

Open Source Projects

Natural Language Processing & Machine Learning Web App

[+0000bjeca Objeca] - [+0000bjeca Objeca]

Developed language-based RNN to analyze text. Designed, implemented, and deployed full-stack application running the model at scale. Database, authentication, task scheduling, unit testing.

Senior Design: 3D Imaging Burn Clinic Device

[+0000bjeca Objeca] - [+0000bjeca Objeca]

Developed hardware and software for point of care 3D imaging of burn patients to more accurately measure TBSA. Validated error of 1.4% (vs. 20% typical).

  • Presented poster @ UC Davis School of Engineering Design Showcase 2019
  • Presented project @ UC Davis Biomedical Engineering Design Symposium, 2019
  • Abstract Published @ American Burn Association Conference, 2020
  • Abstract Published @ IEEE HI-POCT conference, 2019, awarded 'Rising Star in Healthcare Innovations'

Education

Vanderbilt University Medical Center

[+0000bjeca Objeca] - [+0000bjeca Objeca]

PhD Biomedical Informatics

Vanderbilt University Medical Center

[+0000bjeca Objeca] - [+0000bjeca Objeca]

MS Biomedical Informatics

UC Davis

[+0000bjeca Objeca] - [+0000bjeca Objeca]

BS Biomedical Engineering

Awards

Rising Star in Healthcare Innovations Award

IEEE HI-POCT conference

Senior Design: 3D Imaging Burn Clinic Device

Finalist: Student Paper Competition

AMIA 2021 conference

Predicting Motor Responsiveness to Deep Brain Stimulation with Machine Learning

Student Innovation Award

AMIA 2021 conference, Knowledge Discovery & Data Mining working group

Predicting Motor Responsiveness to Deep Brain Stimulation with Machine Learning

Publications

Effects of Image Quantity and Image Source Variation on Machine Learning Histology Differential Diagnosis Models

[+0000bjeca Objeca]
Journal of Pathology Informatics

CNN-based educational tool for generating histology ddx lists from histo whole slide images

Predicting Motor Responsiveness to Deep Brain Stimulation with Machine Learning

[+0000bjeca Objeca]
AMIA Annual Symposium 2021

ML tool to predict effect of deep brain stimulation in Parkinson's patients

Exploring Risk Factors in Suicidal Ideation and Attempt Concept Cooccurrence Networks

[+0000bjeca Objeca]
AMIA Annual Symposium 2022

Used NLP & Network analysis to mine and graphically analyze clinical-note-derived risk factors as they relate to suicidal thoughts and behaviors

Enhancing Suicide Risk Prediction Models with Temporal Clinical Note Features and NLP Techniques

[+0000bjeca Objeca]
Vanderbilt University Medical Center

In progress, Dissertation Research - Using NLP, ML, RF, LSTM to improve suicide risk prediction w clinical notes & temporality

Exploring Natural Language Processing and Sentiment Analysis of Patient Portal Messages for Machine Learning Suicide Risk Prediction

[+0000bjeca Objeca]
Vanderbilt University Medical Center

In progress, Dissertation Research - Using NLP, ML, LSTM, Sentiment Analysis to improve suicide risk prediction w patient portal messages

Completeness and Readability of GPT-4 Generated Multilingual Discharge Instructions in the Pediatric Emergency Department

[+0000bjeca Objeca]
Journal of the American Medical Informatics Association

Using GPT-4 LLM to generate discharge summaries in English and Spanish, and at different reading levels, then analyzing generations for completeness and readability to assess the practicality of the tool.