Machine Learning Research Engineer

WindBorne Systems•Palo Alto, CA

1d•$140,000 - $240,000•Hybrid

About The Position

WindBorne Systems is seeking a broadly capable ML researcher and engineer passionate about machine learning to advance their AI weather models. These models run 24/7, producing global forecasts every 20 minutes, and present rich machine learning challenges due to enormous, messy datasets, imperfect ground truth, complex physical structure, and the need for reliable real-world performance. The role involves pushing models forward, with a focus on identifying important problems and making progress on them, rather than fitting a specific specialty. The company utilizes a proprietary data source: a global constellation of smart weather balloons that capture critical atmospheric data. WindBorne designs, manufactures, and operates these balloons to generate unique weather intelligence. Their mission is to reduce weather uncertainty and aid climate change adaptation, including hurricane prediction and accelerating renewable energy adoption. The founding team, composed of Stanford engineers, was recognized by Forbes 30 Under 30 and is supported by prominent investors like Khosla Ventures and Footwork VC.

Requirements

Strong research taste: ability to identify important questions, design experiments, and recognize valid results.
Deep enthusiasm for machine learning and a desire to understand models thoroughly.
Strong Python and PyTorch skills, with experience developing and debugging non-trivial ML systems.
Experience working with large, messy datasets and building dependable pipelines or abstractions.
Ability to iterate quickly while systematically considering which work should become durable infrastructure.
Comfortable crossing boundaries between research and engineering, and learning new tools or domains as needed.
Experience with weather, climate, geospatial data, scientific machine learning, computer vision, or time-series forecasting is helpful, but not required.

Responsibilities

AI-based data assimilation: Develop methods for incorporating real-time observations from balloons, satellites, weather stations, and other sources into forecasts.
Foundation model for weather: Work towards a single model capable of predicting various weather-related datasets, including variables and data products beyond traditional global forecasts.
Messy, large weather datasets: Find, understand, clean, and combine large datasets with inconsistent formats, resolutions, coverage, and quality. Identify useful data and build systems to facilitate its reuse.
Rapid experiments: Test ideas quickly, learn from failures, and pursue promising results to achieve incremental model improvements.
Infrastructure and systems: Transform successful experiments into reusable systems to enhance future research.
Research direction: Form hypotheses, design experiments, stay updated on relevant ML research, and contribute to deciding which ideas to pursue.
Address operational needs: Contribute to operations, infrastructure, evaluation, data engineering, or other technical areas as required to ensure research success in practice.