Wildfire ML Scientist - Remote Sensing & Field Modeling

Automatic Data Processing, Inc.; Thermal (CA); Remote; USD 140,000 - 210,000; Full time

Job summary

EmberWorks Sensing seeks a pioneering Fire Behavior Modeling Scientist to own and evolve our modeling suite. You will shape methods, validate results, and deliver outputs to operations teams using airborne thermal data.

You will work with geospatial engineers to build a production pipeline, train models on spatiotemporal data, and define compute needs for training and inference on GPU-enabled cloud infrastructure.

Qualifications

• Deep understanding of fire behavior models and underlying physics/maths.
• Hands-on ML or data-driven modeling on spatiotemporal data.
• Strong remote sensing background with raster imagery experience.
• Proficient in Python geospatial stack and Git.

Responsibilities

• Own fire behavior modeling effort from integration to new methods.
• Build and integrate models into automated production pipelines using airborne thermal data and external inputs.
• Research and develop new modeling approaches with empirical, physical, and ML methods.
• Define validation criteria and interpret outputs, diagnosing data/model issues.
• Specify compute requirements for training/inference and coordinate GPU/distributed resources.
• Engage with academic partners and maintain external collaborations.

Skills

Fire behavior modeling (ELMFIRE/WRF-SI

Machine learning on spatiotemporal I

Remote sensing & raster imagery

Python data stack (xarray, GDAL, ras

GPU training on cloud

Independent team member in R&D

Education

Bachelor's degree in fire science, atmospheric science, remote sensing, CS, environmental science or related field

Tools

PyTorch

GDAL

rasterio

Git

Job description

EmberWorks Sensing seeks a pioneering Fire Behavior Modeling Scientist to own and evolve our modeling suite. You will shape methods, validate results, and deliver outputs to operations teams using airborne thermal data.

You will work with geospatial engineers to build a production pipeline, train models on spatiotemporal data, and define compute needs for training and inference on GPU-enabled cloud infrastructure.