Machine Learning Research Engineer, Model Evaluation

About WindBorne Systems

WindBorne Systems revolutionizes weather forecasting by utilizing a unique data source: a global network of next-generation smart weather balloons that collect crucial atmospheric data. The company designs, manufactures, and operates these balloons, enabling the generation of unparalleled weather intelligence. Their goal is to reduce uncertainty in weather predictions, helping humanity adapt to climate change through better hurricane forecasts and accelerating renewable energy adoption. Founded by Stanford engineers recognized by Forbes 30 Under 30, WindBorne is supported by leading investors such as Khosla Ventures and Footwork VC.

Role Overview

The company develops AI-driven weather models that operate continuously, delivering global forecasts every 20 minutes. Evaluating these models poses complex challenges, as performance varies by region, forecast lead time, weather conditions, and customer requirements. Standard metrics often overlook factors that make forecasts meteorologically accurate or practically valuable.

This role calls for a candidate with strong scientific discernment to identify the models' strengths and weaknesses and gauge trustworthy outcomes. The role bridges machine learning and meteorology, combining rapid analytical feedback with dependable systems to accelerate and improve research reproducibility.

Key Responsibilities

• Design and implement a comprehensive evaluation approach in collaboration with the Meteorology team to compare WindBorne's WeatherMesh model against competing AI and physics-based weather models, selecting appropriate metrics, datasets, baselines, and illustrative case studies to accurately assess forecast quality.
• Create fast evaluation processes that provide prompt, actionable insights for researchers, and develop these into durable, reusable systems promoting reproducibility and traceability within the research workflow.
• Enhance existing evaluation platforms, including AI-driven agentic tools that assist in forecast analysis and synthesis of results.
• Produce clear, informative communications such as scorecards, visualizations, and detailed reports for internal teams, leadership, clients, and partners, presenting model performance transparently along with associated uncertainties and important caveats.

Required Qualifications

• Exceptional scientific judgement paired with a healthy skepticism, consistently questioning fairness of comparisons, considering alternative explanations, and demanding solid evidence.
• Strong experimental design skills to identify the most relevant evaluations and confidently distinguish meaningful improvements from statistical noise.
• Ability to think systemically, solving immediate problems while designing evaluation tools and infrastructure to be reused in future projects.
• Proven experience assessing machine learning systems with large-scale, scientific datasets that may be geospatial, multidimensional, or temporal in nature.
• Proficiency in Python and scientific computing libraries such as PyTorch, NumPy, pandas, and xarray.
• Ability to independently analyze ambiguous results, synthesize insights, and communicate conclusions effectively.
• Familiarity with weather, climate, forecasting, physical sciences, or AI research tools is advantageous but not mandatory.

Benefits

• 401(k) retirement plan
• Comprehensive dental, health, and vision insurance
• Unlimited paid time off
• Stock option plan participation
• Office amenities including food and beverages

Compensation & Location

The position offers a competitive salary range from $140,000 to $240,000 annually, adjusted based on candidates' background and experience. Work is conducted onsite at 1600 Bridge Pkwy, Redwood City, CA, with hybrid or fully in-person arrangements.