SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 170,000 - 300,000 / annual
Hadrian is building autonomous factories to reindustrialize American manufacturing, combining AI, robotics, and advanced software to help aerospace and defense companies manufacture mission-critical systems up to 10x faster and at lower cost. The company has raised $1.37B in Series D funding at a $7.87B valuation and is rapidly expanding manufacturing capabilities across welding, casting, forging, electronics, and additive manufacturing.
This role focuses on the modeling and prediction side of manufacturing data science. The core challenge: predict what a manufacturing process will produce before it runs, using rich process data from high-mix, low-volume aerospace parts. Since most parts are near-unique, the traditional "lots of history per SKU" approach doesn't apply. Instead, the leverage comes from representation learning—embedding parts by geometry, material, tolerances, and route to predict behavior from similar parts.
Key responsibilities include building and shipping production models for cycle time, tool life, quality, and demand with calibrated uncertainty (quantile, conformal, or Bayesian methods). You'll engineer features and build geometric/graph models that predict cycle time, cost, DFM, and tolerance risk directly from CAD, mesh, and point cloud data. You'll develop a part and operation embedding layer that represents parts by their characteristics, retrieves similar parts, and transfers their behavior to cold-start new ones.
Validation is critical: you'll backtest rigorously while respecting time ordering and preventing part-family leakage, and make defensible cases for deep versus classical methods. You'll own models end-to-end on the platform—reproducible training, serving, monitoring, and retraining—in partnership with ML Platform and Data Engineering teams. You'll close the loop by detecting drift and quality anomalies so predictions improve as new data arrives, and translate predictions into decisions for quoting, scheduling, capacity, and DFM through experiments and A/B tests.
Required skills include forecasting and prediction on real, messy manufacturing data with honest uncertainty quantification; representation learning, embeddings, similarity, and retrieval; deep learning (PyTorch) with judgment about when to use it; strong classical ML and statistics (GBMs, Bayesian/hierarchical, survival, causal); rigorous validation practices; Python for feature engineering and model deployment; and experience deploying and monitoring models in production.
Differentiators include geometric deep learning (mesh/point-cloud networks, GNNs, PyTorch Geometric), CAD/B-rep knowledge and feature recognition, retrieval and ANN at scale, Bayesian and hierarchical modeling for small data, survival and reliability modeling, aerospace or precision-manufacturing background, digital twins and simulation, and causal inference experience.
