Connectionists: Deep Learning for video, Toronto

[Roland Memisevic]

Deep Learning Research Scientist (Toronto)

Twenty Billion Neurons (www.twentybn.com) is a deep learning startup based
in Berlin (Germany) and Toronto (Canada). Our mission it is to let machines
perceive and understand the world. We train neural networks on videos to
make fine-grained predictions about actions, objects and the state of the
world. This allows us to solve video analysis use-cases, while improving
the understanding of scenes and situations and providing visual grounding
for AI. We created our own video crowdsourcing platform to collect very
large real-world datasets for learning both generic and task-specific video
features. We will have several research openings in our Toronto lab.

Tasks

   -

   Explore neural network architectures for video understanding and making
   detailed predictions about visual scenes.
   -

   Perform research on recurrent networks, attention policies, transfer
   learning and scaling up training.
   -

   Participate in the development of real-time inference methods.
   -

   Publicise your work through blog-posts and/or conferences or journals
   such as ICLR, NIPS, CVPR, PAMI.


We also occasionally co-supervise thesis and project work associated with
university programs (MSc/PhD/Postdoc). If interested, please send us an
informal inquiry at the email address below.

Requirements

   -

   PhD or MSc in computer science or a related area with a strong focus on
   Deep Learning.
   -

   Very good knowledge of Python, and experience with Tensorflow, Theano or
   a similar DL framework.
   -

   Strong programming skills and knowledge of (or at least interest in)
   software engineering practices for working in teams.


To apply, send your resume or CV, transcripts, and any other information
you consider relevant (ideally in a single pdf-file) to:
careers at twentybn.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20170219/31c21a8a/attachment.html>