Abby Van Soest

Intelligent and Ethical Systems

About Me

I am a senior at Princeton University, where I study computer science with a focus on reinforcement learning. I am fascinated by everything from robotics to the history and philosophy of technology. I'm originally from a rural area outside Ithaca NY, where my parents operate a bed and breakfast out of our home.

Interests

Reinforcement Learning

My favorite topic in computer science is reinforcement learning. I want to explore how intuition about general human intelligence can be translated to formal mathematics and implemented on machines.

Backpacking

I fell in love with the backcountry during my gap year, when I solo-backpacked through New Zealand for 3 months. Now, I try to go hiking and backpacking in the mountains whenever possible.

Climate Science

I believe technology can provide a safe, ethical, and scalable response to climate change. My work in ML is in service of socially-motivated missions such as this.

Robotics

I first became interested in robots during a summer research internship at Oregon State University. I am especially passionate about encouraging female participation in robotics.

Skills and Experience

Projects

Reinforcement Learning without Rewards

Collaborative work accepted to ICML '19. I implemented and experimentally evaluated the MaxEnt algorithm on a variety of control tasks. Code is publically available on Github.

Gender Represenations across Country-Specific Word Embeddings

Used NLP tools and statistical analysis to examine the representation of gender in localized areas. Trained word embeddings by applying the word2vec algorithm to 22 country-specific English corpora. I then developed metrics to compare the strength of implicit and explicit gender stereotypes between these countries. Findings indicated a regional correlation between both sorts of stereotypes.

This Game Belongs to Earth

Short found-footage documentary about the use of video games in the development of solutions for climate change and modern climate science education.

Multi-Agent Coordination

Design and implement an experimental simulation in which multiple robots tackle two entangled problems in multi-agent learning, direct collaborate and effective use of communication. Implemented using neuroevolutionary and reinforcement learning techniques.

Course Recommender System

A recommender system for Princeton courses based on semantic and lexicographic analysis. Implemented using natural language processing techniques to extract information from public course descriptions.

Work

Google AI Research Intern

Research algorithms for reinforcement learning in the absence of rewards. Evaluate these algorithms in simulated robotic environments.

February 2019 - June 2019

Google Software Engineering Intern

Worked for two summers within the Google Search division. My team focused broadly on 'News on Search', or the way in which news stories and current events are surfaced to search users.

My first project was to develop and implement a triggering model that decided when to activate a new feature for a Search Engine user.

My second project was examine methods for selecting concise and representative phrases for news stories.

June 2017 - September 2017 + April 2018 - August 2018

Undergraduate research assistant

Research topics in multi-agent learning and coordination under Professor Kagan Tumer at Oregon State University.

May 2016 - August 2016

McGraw Center Math Tutor

Tutor students in calculus and linear algebra through Princeton's center for teaching and learning.

May 2016 - present

The Fontainebleau Inn

Provide friendly, high-quality customer service to diverse groups of guests at weddings and other formal events as a bartender and bar manager.

May 2009 - September 2015

Extras

Overland Leader

Lead high-school aged students on backpacking and sea kayaking expedition in Alaska. Seek to create a positive, educational and resonant life experience for teens in the wilderness.

June 2019 - August 2019

Student Representative on the Committee on Discipline

Work with teachers, administrators, and other students to administer the University's rules and regulations governing academic integrity and serious non-academic misconduct.

May 2016 - present

Residential College Advisor

Work with College staff to mentor and advise incoming freshman students on academics and general life at Princeton.

March 2016 - June 2017

Skills

C and C++
Java
Python
Tensorflow
Optimization
Deep Learning
Writing
Distributed Systems
AWS
HTML + CSS
Git

Contact

Feel free to message me with any questions.