Visiting Researcher
Human Decision Support
Company Description
At ServiceNow, our technology makes the world work for everyone, and our people make it possible. We move fast because the world can’t wait, and we innovate in ways no one else can for our customers and communities. By joining ServiceNow, you are part of an ambitious team of change makers who have a restless curiosity and a drive for ingenuity. We know that your best work happens when you live your best life and share your unique talents, so we do everything we can to make that possible. We dream big together, supporting each other to make our individual and collective dreams come true. The future is ours, and it starts with you. With more than 7,400+ customers, we serve approximately 80% of the Fortune 500, and we’re on the 2021 list of FORTUNE World’s Most Admired Companies®. Learn more on Life at Now blog and hear from our employees about their experiences working at ServiceNow.
Team Description
The ServiceNow Research team does both fundamental and applied research to futureproof AI-powered experiences for all users of the Now Platform®️. We make workflows smarter, AI workloads more efficient, and we are committed to making socially responsible contributions to the AI community. Our innovation is centered on people: helping customers modernize their technology architecture, innovating new business models, improving experiences at work, and driving higher ROI from their technology investments.
The Human Decision Support (HDS) team at ServiceNow Research enables better decisions in a volatile world with tools from causality, time series, and semantic relationships. We are actively seeking collaborators and interns. The topics include but are not limited to:
- Causality
- Time-series forecasting
- Reinforcement learning
- Planning
Job Description
We are looking for a visiting researcher (remote or based in Montreal) to join our team, composed of: Valentina Zantedeschi (ServiceNow), Luca Franceschi (Amazon Web Services), Matt J. Kusner (University College London) and Vlad Niculae (University of Amsterdam). We are working on on learning Directed Acyclic Graphs (DAGs) from observational data, which is a fundamental problem with implications from better modelling to causal discovery.
Instead of modelling directly in the space of DAGs, which is difficult to characterize given its discreteness and most importantly the acyclicity of the set of solutions, we aim to parametrize and optimize over the space of topological layers. Indeed several works, including our recent paper https://arxiv.org/pdf/2301.11898.pdf, decompose the problem of DAG learning into (1) learning a total ordering of the variables (and masking out the edges that are not consistent with it) and (2) pruning unnecessary edges. The rationale is to guarantee the validity of the DAG without enforcing acyclicity constraints and to work on the smaller and more regular space of orderings. However, a DAG (if it is not complete) can be consistent with multiple total orderings, hence it would be preferable to use a parametrization that does not distinguish between equivalent causal orderings, as the one offered by learning topological layers.
As a Visiting Researcher:
- You will conduct research and publish the results at top-tier conferences such as NeurIPS, ICLR, ICML and AISTATS.
- You will get access to all the needed resources, including a cluster with GPU resources.
Qualifications
In order to be successful in this role, we need someone who has:
- B.E./B.Tech./M.E./M.S./M.Tech. from Computer Science, Electronics and Communications, and Electrical Engineering fields with an excellent academic record.
- Currently enrolled in Machine Learning-related M.S. or Ph.D. program.
- Minimum 1+ years of experience in research with a good publications record.
- Experience with deep learning frameworks such as Pytorch or Tensorflow.
Apply Now