Applied Machine Learning Pre-doctoral Research Professional Application

Pre-doctoral fellowships are an opportunity to work directly with faculty and experience research projects first hand for two years, before applying for a PhD. Our pre-doctoral fellows design and implement entire research pipelines, from data processing, to model development and evaluation, to summarizing and presenting results. In addition, fellows are able to connect with visiting experts and external collaborators, attend interactive seminars with University faculty, and participate in joint lab meetings, collaborating with and learning from other research teams.

For more information about pre-doctoral fellowship positions in general, including their benefits, help on applying, and even a thorough listing of other similar opportunities, we recommend visiting predoc.org. For more information about a pre-doctoral fellowship at the Chicago Booth School of Business, refer to the information session link above.

We are a research center at the University of Chicago Booth School of Business. Our research lab includes principal researchers, pre-doctoral fellows (one of which could be you), part-time research assistants, designers and developers: you can find members of our team here. We use our diverse skills to create, publish and share some of the best and most exciting research in economics and social science. We value academic excellence, high-quality academic research and the ability to improve the world around us.

You can read more about the team on our website:
https://www.chicagobooth.edu/caai.

Our research focuses on using machine learning with impact. This means taking some of the most sophisticated tools from computer science, statistics, machine learning and AI, and applying them to a range of challenging problems. Here are some problems our pre-doctoral fellows have worked on:

• We use machine learning to study the choices of medical professionals deciding whether to test for heart attacks in their patients. This work identified two inefficiencies: predictably low-risk patients are tested with little benefit, while predictably high-risk patients are left untested, and suffer adverse health events. By comparing algorithmic predictions to the decisions made by doctors, we can suggest several sources of error, which you can read about in this recent paper in the Quarterly Journal of Economics.
• We apply a notoriously biased algorithm, facial recognition, for a good purpose. Human decision-makers can rely on faces, even when they ought not to. Using data collected by the CAAI, we trained a face-algorithm that uncovers biases in how people make such choices. You can see an early recorded presentation of this work on YouTube.
• Our recent paper in Science studied a widely-used health care algorithm that affects decisions on nearly a hundred million patients in the US. We found this algorithm had significant racial bias - Black patients were rated as lower risk than equally healthy White patients. We have now begun creating a fix for this algorithm and are working with systems to implement fixes.

Here are some researchers we’ve worked with in the past or are currently working with:

• Jens Ludwig, UChicago Crime Lab
• Ziad Obermeyer, Berkeley School of Public Health
• Jon Kleinberg, Cornell University
• Kate Baicker, Harris School of Public Policy
• Jure Leskovec, Stanford University
• Mukund Sundararajan, Google
• Emma Pierson, Cornell Tech
• Himabindu Lakkaraju, Harvard Business School
• Devin Pope, Chicago Booth
• Oeindrila Dube, Harris School of Public Policy

We are looking for candidates with a strong academic record, and an interest in pursuing research excellence in the future.

Candidates must have:

• A bachelor’s or master’s degree in computer science, statistics, data science, economics or a closely related field. Applicants must have either graduated from such a program, or intend to graduate by mid 2022.
• Some interest in a PhD, enough to invest in or explore that interest, whether in economics, social science, data science, computer science, or another related field.

Candidates should excel in one of these skills, and want to learn about the others:

• Data science, mathematics, or statistics: do you know about a variety of models such as linear regressions, binary trees, or convolutional neural nets? Have you built a pipeline that tackled a big hairy dataset? Do you know when to use splines or regularization?
• Economics or social science: do you know what a regression discontinuity is? Have you had experience from a project or class applying economic ideas and seeing the subtleties of social data?
• Computer science: do you have some expertise in either a single methodology (e.g. convolutional nets or generative models) or a single modality (such as images or language)? Have you designed custom architectures for to solve complex real-world machine learning problems?

Experience training deep learning models is valued, but isn’t required to apply. Many of our pre-doctoral fellows have developed their model training skills since joining the team. If learning how to train a CNN sounds interesting, and you think you would be able to develop the skill, then we encourage you to apply.

Our most recent pre-doctoral fellows have secured PhD positions at Harvard, Berkeley, Carnegie Mellon, and here at the University of Chicago.

We value diverse perspectives and experiences, and we especially encourage those from non-traditional backgrounds to apply.

We have sponsored Visas in the past and encourage international applicants to apply. We evaluate all candidates equally regardless of where they are applying from.

Yes, this position is in-person hybrid. Successful applicants are expected to join the team on the University of Chicago campus in Hyde Park and follow the current work location policies outlined by Chicago Booth.

We recommend applying for the first round of hiring if possible. If you miss the first deadline, we will consider your application for the spring hiring round. If you’d like to be considered independently of these hiring periods, please reach out directly.

We do our best to accommodate every applicant. If you need some variation to the application timeline, simply email with your name and request and we will do our best to accommodate you.

Successful applicants will generally start around July 2023. Earlier and later start times can be easily facilitated.