Upstart's Decisioning org is forming a new, high-leverage Applied Machine Learning team to push the boundaries of model accuracy in their underwriting systems. This team will serve as the applied ML counterpart to their centralized ML Science group, chartered to drive model precision by focusing on feature engineering, model tuning, embedding optimization, and CUDA-accelerated training workflows. The goal is to drive improvements that have direct business impact on pricing accuracy and borrower conversion.
Requirements
- Proficiency in Python and core ML frameworks (e.g., PyTorch, TensorFlow, Scikit-learn, XGBoost).
- Demonstrated expertise in end-to-end model development: data prep, feature engineering, training, evaluation, and deployment.
- Practical experience optimizing ML workflows using CUDA/GPU acceleration.
- Strong grasp of regression and classification metrics (e.g., precision, recall, R², MPVRMSE) and how to apply them to production models.
- Background in feature store design, embedding architecture, or synthetic data generation for model training.
- Familiarity with modern experimentation frameworks, hyperparameter tuning tools, and automated model selection techniques.
Responsibilities
- Serve as the technical lead for applied ML initiatives that improve the accuracy, precision, and recall of underwriting models.
- Design and implement advanced ML training strategies, including AutoML, ensemble learning, and temporal modeling techniques.
- Drive GPU-accelerated experimentation, including CUDA-based training optimization and embedding fine-tuning.
- Build robust data preprocessing and feature engineering pipelines that can be used in both experimentation and production.
- Influence modeling strategy through close collaboration with Pricing Engineering and the ML Science organization.
- Deliver measurable improvements to model-driven business outcomes such as conversion rate, rate accuracy, and loan performance.
- Mentor future applied ML engineers and help define the long-term roadmap for ML excellence within Pricing.
Other
- 8+ years of hands-on experience in applied machine learning, with strong exposure to production-scale modeling efforts.
- Ability to work autonomously and lead technical direction in ambiguous, high-impact domains.
- Experience working in high-scale, ML-driven product environments—especially in fintech, pricing, or risk modeling.
- Proven track record of improving model accuracy in production environments with measurable business outcomes.
- Ability to bridge engineering and science teams, and influence technical strategy across disciplines.
