OverviewResponsibilities
Duties of the Scientific Developer for AI/ML Forecasting will include the following:
- Conduct cutting-edge development and innovation in next-generation forecasting models and technologies driven by Super Intelligence/Artificial Intelligence/Machine Learning (SI/AI/ML).
- Serve as a key technical resource for SI/AI/ML-based modeling suites (including AIGFS, AIGEFS, and HGEFS).
- Identify, prototype, evaluate, and refine emerging machine learning approaches; conduct numerical experiments, analyze performance results, and apply parallelization and code optimization to develop successful experimental models for operational implementation.
- Establish best practices, standards, and scalable CPU- and GPU-based workflows for training and inference across high-performance computing (HPC) and cloud environments.
- Collaborate with cross-functional engineering teams to build, optimize, and maintain efficient data-loading pipelines handling diverse meteorological data types.
- Integrate SI/AI/ML tools, models, and workflows directly into Office of Modeling and Development (OMD) modeling suites.
- Collaborate with the OMD team, partners, and the broader research community to advance state-of-the-art weather and climate forecasting through data-driven approaches and innovative machine learning methodologies.
- Stay current on the latest scientific, methodological, and technological advances in SI/AI/ML-driven weather forecasting.
- Promote software engineering and research best practices through regular code reviews, documentation, and maintenance of high code quality standards.
- Train and mentor scientists on machine learning capabilities, AI tools, and frameworks, and participate in technical workshops and professional conferences to communicate findings.
Qualifications
The Scientific Developer for AI/ML Forecasting selected should have the following:
- Background in machine learning/data science, meteorology/NWP, physical sciences, mathematics, computer science, scientific computing, or a closely related field with demonstrated SI/AI/ML-related or closely related computing experience.
- Demonstrated proficiency with Python programming and work experience in ML frameworks such as PyTorch or JAX, coding/scripting, and automated workflow development (Bash, Python).
- Strong working knowledge of core ML lifecycle principles, including data wrangling, model training, validation, and operationalization, as well as architectures including Graph Neural Networks (GNNs), Transformers, and Diffusion Models.
- Demonstrated experience or working understanding of Earth systems data, modeling, and workflows, or other scientific computing experience.
- Experience leveraging SI/AI tools (LLMs, agents) to advance code and documentation development, or willingness to learn.
- Demonstrated ability to independently identify, prioritize, and implement bug fixes and new features.
- Working knowledge of version control, collaborative code management, automated code testing, and other modern software engineering practices.
- Ability to work effectively and professionally as part of a diverse, customer-focused team, collaborating with physical scientists, software engineers, and inter-agency partner teams.
- Good verbal and written communication skills in English.
- Must be a U.S. Citizen or Green Card Holder.
The Ideal Scientific Developer for AI/ML Forecasting will have the following:
- Demonstrated experience with SI/AI/ML-driven weather forecasting methodologies, including bias correction, full-SI/AI/ML forecasts, physics-informed neural nets, etc.
- Familiarity with SI/AI/ML forecast models and frameworks (Google DeepMind, WeatherNext, Earth-2, Anemoi), as well as training data sets (ERA5, GDAS/GFS).
- Experience with code optimization, including for GPU and other parallel computing.
- Experience with containerization (Docker, Podman, Singularity/Apptainer), continuous integration (GitHub Actions, GitLab CI, Jenkins), and DevOps and Agile methodologies.
- Build systems and software dependency management (pip, Conda, Spack).