Columbia University in the City of New York (CU)

Staff Associate

Mortimer B. Zuckerman Mind Brain Behavior Institute

New York, New York, NY, United States

Columbia University’s Mortimer B. Zuckerman Mind Brain Behavior Institute (the Zuckerman Institute) brings together researchers to explore aspects of mind and brain, through the exchange of ideas and active collaboration. The Zuckerman Institute’s home will be the Jerome L. Greene Science Center on Columbia’s new Manhattanville campus. Situated in the heart of Manhattan, at full capacity the Zuckerman Institute will house approximately 47 laboratories employing a broad range of interdisciplinary approaches to transform our understanding of the mind and brain. In this highly collaborative environment, labs work together to gain critical insights into human health by exploring how the brain develops, performs, endures and recovers from trauma or disease.

The Kriegeskorte Lab within the Zuckerman Institute is currently seeking a Staff Associate. Animals can behave successfully in a complex world that is only partially predictable to them. To achieve this they must learn (1) what latent variables of their environment to devote their representational resources to, (2) how to represent those latent variables in neural activity, and (3) how to predict the future state of these latent variables.

The Staff Associate will (1) develop a range of simulated dynamic environments of varying complexity to train and test candidate models with, (2) develop a range of cost functions and neural network model architectures that implement the unsupervised learning objective, (3) to train and test the architectures and cost functions in the simulated environments, and (4) to write and submit a paper on these developments to a journal and/or engineering conference.

This project is motivated by the hypothesis that animals learn a “predictable scaffold of reality” that enables them to ignore what is unpredictable about their environments and exploit for action and planning what is predictable. The goal of this position is to develop a computational model for unsupervised learning of an informational reduced representation of dynamic visual input that prioritizes representation of latent variables that are either predictive or predictable.


Bachelor’s degree

One year of related experience. High level of programming expertise is required.

A strong background in mathematics, statistics, neuroscience or related field is highly desired.

Please apply via recruiter’s website.