When I was a graduate student at UCLA, I had the opportunity to see Benoit Mandelbrot speak. My biggest takeaway from the talk was not about the fractal dimension of the coastline of England, but rather the importance of learning everything possible, even if it doesn’t seem immediately applicable to our own research. Everything we learn is a “tool in our toolbox.” All of our seemingly disjoint classes from graduate school, lectures we have heard, and random articles we have read may not especially seem relevant at the time, but may help us later when we need a new way of thinking about our problem. The Toolbox Philosophy has proven true for me time and again in my own research, and often motivates me to work in problems that are somewhat outside my own domain, since I am then challenged to see data differently and learn new techniques.

I am the founder of
ChatterBaby, creating an algorithm to translate a baby's cries into labels (cry vs. no-cry, pain vs. fussy vs. hungry) using machine learning on acoustic features. I began this project in 2013 after the birth of my third child, when I realized that my new baby had very similar cry features to her siblings. Five years later, we are releasing the app and launching our ChatterBaby NICU study, where we are creating a "pain thermometer" by predicting the pain levels contained in an infant's cry.

I try to keep a "pet project" going at all times, to explore new techniques and fields. My past side project was computer aided screening of diabetes using the electronic health record. Although diabetes is the fifth leading cause of death in the United States, roughly 1 in 4 diabetics (type 2) are unaware they have the disease. We found many interesting predictors for diabetes, including
ICD 302.X, "Sexual and Gender Identity Disorder."

A statistician is someone who tells you, when you've got your head in the fridge and your feet in the oven, that you're – on average - very comfortable.

Pasted Graphic

Screen Shot 2016-03-17 at 10.30.01 PM