About

I'm a Seattle based scientist. I am currently building NewLimit to develop new therapies based on partial reprogramming.

Before that, I was at AWS working new and diverse technologies to make the world a better place.

Before that, I was at the Allen Institute for Cell Science where I built predicitve models to estimate the outcome of new experiments and describe how cells change their organization under different conditions.

Before that, I was in grad school at Carnegie Mellon University as a student of Robert Murphy. My work focused on applying generative models to determine how cells respond to perturbations.

Even before that, I developed regulated document management software at Amgen.

Undisclosed project 2

Amazon - Special Projects • 2021 - 2022

Sequencing engineering via machine learning and language models for biotechnology applications.

Undisclosed project 1

Amazon - Special Projects • 2020 - 2021

Technical lead on a "speedboat" from PRFAQ to full-initiative funding for entirely new projects at Amazon. Demonstrated the technical feasibility of a new products and services. Contributed to projects in audio, video, and media applications.
Amongst other things, this project resulted in a patent for facial pose retargeting.

Label-free prediction of three-dimensional fluorescence images from transmitted light microscopy

Allen Institute for Cell Science • 2016 - 2020

With collaborators from the Allen Institute for Brain Science, I've been developing tools to determine relationships across image modalities. We use a very simple deep learning model to predict the organization of many types of subcellular structures from three-dimensional brightfield images. We also show that this method works for electron micrographs, allowing us to solve very difficult image registration problems.

You can read our manuscript from September 2018 at Nature Methods.
The software to build and run our models is here.

Integrated Cell Models

Allen Institute for Cell Science • 2016 - 2022

I've been working in collaboration with Rory Donovan-Maiye to build integrated models of single cells. We use some cool deep learning tricks and conditional generative adversarial networks (GANs) to fuse data from multiple fluorescence microscopy experiments into a coherent model of sub-cellular structure localization in single cells.

Our manuscript for 3D models can be found in PLOS Computational Biology.
Our software can be found here.

Dose-response screening in the low-replicate high-content regime

Carnegie Mellon University• 2015 - 2017

To efficiently determine the effects of perturbations on cells (drugs), we need to build quantitative automated analysis tools to analyze the results of experiments. With collaborators at the University of Freiburg, we designed a pipeline to detect dose-dependent responses in plant cells in experiments with low numbers of replicates and high experimental variation. By labeling different components, we learn when and where these drugs affect the cell.

Our paper can be found in Cytometry.

Characterization of Starvation-Induced Macroautophagy

Carnegie Mellon University• 2014 - 2017

When cells get hungry they eat themselves. With different collaborators at the University of Freiburg, we showed that depending on how they get hungry they eat themselves in different ways. We characterized how, when, and where this happens. Our publication made it on the cover of the journal Autophagy.

The manuscript can be found in the journal Autophagy.

CellOrganizer

Carnegie Mellon University• 2012 - 2016

A large part of my Ph.D. was working on CellOrganizer, a generative model that describes how cells are put together. That resulted in a few projects:

A non-parametric model of how cell and nuclear shape are coordinated across cell states and drug conditions was published in Molecular Biology of the Cell.

Two models of how compartments of the cell are organized with the cell superstructure were published in PLOS Computational Biology and Cytometry.