Skip to main content

This job has expired

Machine Learning Scientist

Employer
NCAR
Location
Boulder, Colorado
Closing date
Jan 4, 2020

View more

Discipline
Atmospheric Sciences
Career Level
Early Career (up to 10 years past degree)
Education Level
PhD
Job Type
Full-time
Relocation Cost
Paid
Sector Type
Non-Government Organization/Non-Profit

Job Details

Where You Will Work:

NCAR’s Computational and Information Systems Laboratory (CISL) is a leader in supercomputing and data services necessary for the advancement of atmospheric and geospace science. CISL’s mission is to remain a leader at the forefront of ensuring that research universities, NCAR, and the larger geosciences community have access to the computational resources they need for their research. To fulfill the need for a stronger workforce at the intersection of High Performance Computing (HPC) and geoscience problems, CISL engages in education and outreach activities to inspire and attract a diverse future workforce.

What You Will Do:

The Machine Learning (ML) scientist will apply ML to the Earth system sciences as part of the Analytics and Integrative Machine Learning (AIML) Group in the Technology Development Division (TDD) in the Computational and Information Systems Laboratory (CISL) at the National Center for Atmospheric Research (NCAR). The incumbent will have a proven ability to apply a variety of ML algorithmic approaches to problems in earth systems science or related physical science disciplines.

Relevant ML application experience may include the use of ML training and inference systems for the recognition, prediction or tracking of important features or events in datasets, or alternatively, through the auto-encoding of suitable physics parameterizations in earth system models with neural networks, the replacement of model components with efficient, learned emulators. Machine learning techniques may also be applied by the  scientist to automate or accelerate the human data analysis of hundreds of routine data products, thus amplifying the scientific capability of researchers, the integration of non-traditional data sources into earth system prediction systems, to help optimize supercomputing workflows through ML-guided resource management, or for the early detection and steering of numerical simulations.

 

                                                                                                                                                                          

Responsibilities:

The position works on machine learning projects with scientists, engineers and students in NCAR’s Computational and Information Systems Laboratory (CISL), the Research Applications Laboratory (RAL), the Earth Observing Laboratory (EOL), the High Altitude Observatory (HAO), and the Atmospheric Chemistry, Observations and Modeling (ACOM) Laboratory, and potentially with external data scientists as well. The successful candidate will identify and apply appropriate machine learning techniques focused on two initial projects:

  • Emulating atmospheric organic chemistry reaction networks. Current chemistry-climate models cannot represent the complex chemistry involved in the degradation of hydrocarbons emitted by anthropogenic activities, due to the enormous number of species and reactions. ML emulation may produce less costly reduced models, providing new opportunities to investigate their impact on human health and air quality.
  • Data bottlenecks in cloud observational systems. Estimating the radiation balance in the Earth system is central to predictive climate models and hinges on understanding cloud processes.  The Holographic Detector for Clouds (HOLODEC) is an airborne instrument that gives an unrivaled view of 3-D distributions of droplets, providing an unprecedented accuracy and detail of cloud physics data. However, analyzing the huge data volumes produced by the instrument with current techniques presents a bottleneck, limiting the instrument’s scientific utility. Image-based ML methods could accelerate the analysis process and help advance our understanding of cloud processes. 

For each project, the scientist will work with the team, and in collaboration with domain scientists, to create and share the necessary training datasets, and apply, tune and evaluate and verify a variety of machine learning approaches to solving these problems. The machine learning scientist’s efforts will be built on top of NCAR’s core capability in domain-focused statistical development, and will leverage its vast observational and model output datasets, and CISL’s petascale supercomputing infrastructure, and cloud-based resources and environments. The position will require the ability to work in teams and across disciplines in order to cross-fertilize ideas and build strong collaborations to tackle Earth system science challenges. This integration with and support from colleagues in the earth system sciences will help to ensure the relevance and sustainability of the ML project scientist’s research activities.

The position provides high-level machine learning expertise to these projects, assists in planning the projects human and financial resource requirements, and will participate in the evaluation of the project’s progress, its results, and make adjustments to the project’s approach to better achieve objectives. The Machine Learning scientist may also serve, from time to time, as a consultant to internal staff and external organizations on machine learning topics.

  • Communication of Results: The scientist participates in mission-relevant academic activities including conferences, workshops and tutorials. Documents research results by authoring peer-reviewed conference and journal publications, and publicizes those results in presentations at scientific meetings. As a subject matter expert, helps develop grant proposal concepts, teams and text, and may be called upon to serve as a principal or co-principal investigator.
  • Community Service: The position responsibilities will also include an education, training and outreach component. An ML short course developed by the team has now been conducted at a number of venues. The scientist will we help support and further develop this material, as well as initiatives to increase the organization’s training capacity in data-centric science, including machine learning. The scientist serves as a reviewer on scientific papers and proposals, or on conference organizing committees.

 

What You Need:

Knowledge, Skills, and Abilities:

  • Ability to work both independently or collaboratively as a group lead to solve routine and/or occasionally complex technical problems. Ability to organize, prioritize and coordinate multiple tasks. Ability to conduct research with minimal supervision.
  • Excellent communication skills in presenting scientific research, and writing papers in scientific journals, technical reports and proposals. Ability to work and communicate with an international and multi-disciplinary team.
  • Advanced knowledge of one or more machine learning algorithms and the supporting mathematics.
  • Knowledge of one deep learning framework, (e.g. Theano, TensorFlow, Keras), Torch, and familiarity with problem solving environments like Jupyter Notebooks. Knowledge of at least one general machine learning framework (e.g. scikit-learn).
  • Able to use statistical methods to evaluate Machine Learning models.
  • Experience mentoring and working with students.  May supervise the work of others, including project staff. 
  • Awareness of recent developments in the area of computer architecture and high-performance computing. Familiarity with high performance computing environments. Knowledge of CUDA or OpenCL and familiarity with geoscience applications desirable.
  • Solid Python programming skills. Ability to understand and modify code written in C or C++, and Fortran 90.
  • User level familiarity with Linux and Unix-based tools for scripting and file manipulation.
  • May write funding proposals and reports to funding organizations.  May be a PI or Co-Principal Investigator with a member of the Scientific or program staff.
  • May author scientific reports and publications and give presentations at scientific meetings.
  • Represents the organization in providing solutions to difficult technical issues associated with specific projects.
  • May assist in mentoring, supervising, training and/or directing the work of others.
  • Ability to perform occasional/infrequent travel, if required.

Desirable, but not required :

  • Familiarity with distributed/shared memory parallel computing a plus.
  • Familiarity with the use of commercial cloud services will be helpful.
  • Experience in computational earth system science and/or in atmospheric science is desirable.

Education & Years of Experience:

  • Ph.D. in Computer Science or in a physical science discipline which uses machine learning and at least two years of post-graduate experience in the scientific field of specialization; or an equivalent combination of education and experience.

 

Applicant Notes: While the position description details both minimum requirements as well as desired skills and experience, we want to remind applicants that you do not need to have all the desired skills and experience to be considered for this role. If you have the passion for the work along with experience in machine learning, you are encouraged to apply.

 

Company

National Center for Atmospheric Research
Company info
Website

Get job alerts

Create a job alert and receive personalized job recommendations straight to your inbox.

Create alert