Machine Learning Research Engineer, Model Evaluation

Machine Learning Research Engineer, Model Evaluation Machine Learning Research Engineer, Model Evaluation apply now WindBorne Systems is supercharging weather forecasts with a proprietary data source: a global constellation of next-generation smart weather balloons targeting critical atmospheric data. We design, manufacture, and operate our own balloons, using their observations to generate otherwise unattainable weather intelligence. Our mission is to eliminate weather uncertainty and help humanity adapt to climate change—whether by predicting hurricanes or speeding the adoption of renewables. The founding team of Stanford engineers was named Forbes 2019 30 Under 30 and is backed by top-tier investors, including Khosla Ventures and Footwork VC. WindBorne builds AI weather models that run 24/7, producing global forecasts every 20 minutes. Evaluating these models is much harder than producing a headline accuracy number: performance varies across regions, lead times, weather regimes, and customer use cases, while standard metrics often fail to capture what makes a forecast meteorologically sound or useful. We need someone with excellent scientific taste to determine where our models excel, where they fail, and which results we should trust. You will work at the intersection of machine learning and meteorology, combining fast analyses with robust systems that make future research faster and more reliable. ‍ Snapshot of the balloon constellation ON MARCH 31, 2026

Responsibilities Evaluation strategy — Work with our Meteorology team to develop a rigorous, meteorologically valid strategy for comparing WeatherMesh with leading AI and physics-based models. Choose the metrics, datasets, baselines, and case studies that provide an honest picture of forecast quality. Fast feedback and durable systems — Build quick evaluations that give researchers useful signals, then turn recurring analyses into reliable, reusable infrastructure. Think systematically about reproducibility, provenance, and how evaluation tools fit into the broader research workflow. Agentic tooling for evals — Improve our existing evaluation infrastructure, including agentic AI-based tools for investigating forecasts and synthesizing results. Technical communication — Produce clear scorecards, visualizations, and explanations for researchers, leadership, customers, and external partners. Communicate model performance precisely, including uncertainty and important caveats. ‍ Skills and