Andrew Carr

I am a computational mathematician that specializes in machine learning working on my PhD. I'm advised by David Wingate. I'm currently a research intern at Google Brain working with Quentin Berthet on applications of differentiable programming in representation learning. My previous work was in geometric deep learning using the Wasserstein distance and graph processing to improve Neural Processes. I'm broadly interested in optimal transport, generative modeling, and quantum cognitive theory. Feel free to check out my blog or follow me on twitter for random musings about math / deep learning / programming.

twitter linkedin

Google Brain, Paris

This coming summer I am going to join the Brain team in Paris to work on optimal transport and differentiable optimization research.

I'm excited


Lyft, Level 5

I was a machine learning intern on the Prediction team in Lyft's automonous vehicles devision. I worked on dynamic vehicle motion prediction. I used a combination of physics based and data based predictive tools.

I also built a number of tools for my team that increased their development speed. These included a robust A/B platform, visualization tools, build scripts, and other process specific tools


On the side

I have tons of random projects in applied math, machine learning, and data science. Feel free to check out my projects page or my blog.



I worked as a machine learning engineering intern at Qualtrics on their text team. I had two successful projects during this time.

One was for spam detection which saved significant money in potential damages. The other project was a novel topic tracking algorithm. This work, which was a blend of LDA and probabilistic kalman filters, was eventually patented and is a key piece of technology for the team.


Amazon Alexa Prize

I was a member of Team EVE for the Amazon alexa prize challenge. We were 1 of 8 teams selected out of 200+ for this competition.

We were tasked to build a social conversational agent for Alexa that could carry on a 20 minute conversation about arbitrary topics

This problem is still unsolved and research is ongoing.

I built offensive speech filtering tools that beat Amazon's internal tools. I also designed a complex sentiment analysis tool that operated as a heuristic for knowledge retrieval from the web


Competitive Coding Instructor

I was approached by the director of the applied math program and asked if I could teach a class on competitive coding and algorithms. Mitch Probst and I designed the class, implemented curriculum, and taught for 2 semesters.

12 of our students placed in the top 20 of various competitions throughout the year. Many have since gone on to great programming jobs at top organizations.


Private Capital Group

I spent a summer working at a real estate investment firm as a software engineer. I built back office software for the law team that helped save hundreds of thousands of dollars per year. I also built a CRM for the sourcing group.


Development TA

While a student in the applied math program. I worked to write programming labs for the curriculum. We were building the ship as we sailed. I worked on graph theory, optimization, profiling, computer vision, and optimal control theory.

I was also a TA for the control theory course while concurrently enrolled. This was another "trial by fire" experience in learning new material.


Perception, Control, Cognition Lab

I worked with Dr David Wingate for a number of years on fundamental machine learning research. I won a number of competitions, spoke to large groups, presented key findings, and published our work.



I worked as a quality assurance intern at Domo on their internal app store. I reviewed 100s of apps, and wrote several myself. These apps were used by customers to visualize different types of data from various sources.


Carnegie Mellon University

I spent a summer in Pittsburgh as an IT Lab Fellow in the Heinz College. I researched body cameras and police violence along with videogames as a tool to assist refugees learning english.


Math Research

Soon after discovering my love for mathematics, and switching into the applied and computational math program, I started working with Dr Vianey Villamizar on numerical solutions for the wave equation.

I was in over my head, but loved every moment. I took a few grad classes to get up to speed.


BYU Studies

On campus I worked on a VB.NET application. I built a number of custom order websites in a variety of technologies. This work supported me through a large portion of my studies.


Summit Engineering and Consulting

After high school, I worked as a web developer for an engineering firm. I was hired on with very minimal programming knowledge. I went to the local library and checked out every programming book they had. I'm not sure why I thought "Perl in a Nutshell" would help me style my HTML pages.

I made a few web sites for local businesses in the area and for the firm itself. It was an extremely formative time of self study.