Research Software Engineer (Python, HPC, Numpy, Pytorch) - Part-time - Worldwide

Role Overview

We're looking for a skilled Freelance Research Software Engineer. Your role will be to design and implement a reusable experimental platform for environmental machine-learning research. This platform will operate on the University’s Moriah HPC cluster, empowering users to independently configure, execute, monitor, reproduce, and compare experiments.

Responsibilities

• Design and develop a modular, user-friendly research platform for scientific experimentation.
• Create reproducible pipelines for dataset ingestion, validation, preprocessing, harmonization, and caching.
• Develop a configuration-driven interface for dataset selection, representation, and model training.
• Integrate with Slurm for job management, including submission, monitoring, logging, and failure recovery.
• Implement an experiment tracking and management system in collaboration with the research team.
• Ensure automatic recording of experimental configurations, including code versions and metrics.
• Provide a browser-accessible interface for experiment visualization and comparison.
• Develop standard templates for baseline models and evaluation metrics.
• Ensure the platform is maintainable and extensible for future enhancements.
• Deliver comprehensive documentation and training for both users and developers.

Required Skills

• Strong experience in Python software development.
• Proven expertise in building maintainable scientific computing, data engineering, or ML systems.
• Familiarity with Linux and HPC environments.
• Experience with Slurm or similar job schedulers.
• Proficiency in scientific Python tools like NumPy, pandas, xarray, and Dask.
• Knowledge of machine-learning workflows, preferably with PyTorch.
• Experience with experiment tracking tools such as MLflow, ClearML, Weights & Biases, or similar.
• Proficiency in Git, testing, environment management, and software reproducibility.
• Ability to process large multidimensional datasets.

Nice to Have

• Background in hydrology, meteorology, climate science, or Earth observation.
• Experience with geospatial Python tools like GeoPandas, Rasterio, or rioxarray.
• Familiarity with scientific data formats such as NetCDF, Zarr, and Parquet.
• Knowledge of graph and spatiotemporal datasets.
• Experience with Apptainer/Singularity or other reproducible HPC environments.
• Skills in developing software for researchers or non-specialist users.