Pascal Sturmfels

University of Washington ยท PhD Candidate in Machine Learning

I am currently a PhD candidate at the University of Washington, studying machine learning. My previous research includes developing sequencing visualizations from computational biology experiments and novel algorithms for NP-hard scheduling problems. I received my Bachelor's from the University of Michigan in 2017, where I developed novel architectures for convolutional neural networks applied to structural brain images.


Domain Guided CNN Architectures for Brain Imaging

University of Michigan

With the MLD3 lab at the University of Michigan, I developed novel modifications to convolutional neural networks to better understand patterns in the brain. My method resulted in a network that both trained faster and better predicted age than an equivalent CNN baseline. I demonstrated that my methods were more appropriate than standard CNNs in a variety of environments, including settings with small training sets.

September 2017 - May 2018

Software Engineering Intern


I developed a framework in Spark to stress test a client-facing SQL system, as well as a pipeline to more efficiently query gigabytes of financial data on the cloud. I also implemented a modular system that automatically monitored my team's cloud usage, and scaled cloud resources down during periods of inactivity. This reduced our team's expenditure on cloud resources by up to 30%.

May 2017 - July 2017

Mobile Applications Researcher

University of Michigan

I developed a novel application using mesh-networking technology that communicates between phones using peer-to-peer WiFi. The application successfully demonstrated infrastructure-less, censorship-resistant communication.

January 2016 - December 2016

Algorithms Researcher

University of Maryland

I participated in the CAAR REU at the University of Maryland. My group and I demonstrated a novel scheduling framework for NP-hard scheduling problems. This framework improved upon the best known approximation algorithms and has applications in job-scheduling models like MapReduce. I presented the following poster at a research symposium hosted by national science foundation.

June 2016 - August 2016

Computational Biology Researcher

University of California, Berkeley

I worked for a year with the Pachter Lab developing next-generation sequencing software. I developed a system to automatically analyze and visualize results from published papers. This work was published in BMC Bioinformatics, and promotes reproducibility in the field.

June 2015 - May 2016


University of Michigan

Computer Science, Bachelor of Engineering

GPA: 4.00
EECS Scholar
James B. Angell Scholar

Machine Learning
Natural Language Processing
Computer Vision

Instructional Aide, Machine Learning

I taught the foundations of machine learning: support vector machines, linear regression, ensembles and boosting, deep learning, clustering and probabilistic graphical models.

Instructional Aide, Theory of Computation

I covered the foundations of computer science theory: turing machines, decidability and complexity. I also covered standard algorithm paradigms: greedy, divide and conquer, dynamic programming and graph algorithms.

September 2014 - December 2017

Personal Projects

Game Development

Unity, Swift

Apart from doing research in machine learning, I am an active game developer. I am working remaking and revamping an old flash game for the iOS platform, which should be released in January 2018. Although that particular game is done in Swift with SpriteKit, I now do most of my game development in Unity. I enjoy making collaborative multiplayer games and puzzle games, but am not limited to a particular genre.


University of Michigan

I took an active role in the Michigan Student Artificial Intelligence Lab , where I used to lead discussion groups and give presentations about machine learning. Mostly, I used the computer science department's money to order free food for our members.