Staff Data Scientist - Credit and Lending ML

Mercury•San Francisco, NY

About The Position

Mercury is building financial infrastructure for startups and growing businesses and extending credit using proprietary data such as bank transactions, financial reporting, and payroll. The Staff Data Scientist will build and own the inference layer for Mercury’s small business lending program, working closely with Credit, Product, and Engineering. The role includes putting models into production with monitoring and documentation, and helping Credit and Product determine lending decisions and product design. The work spans models for transaction classification, recurring inflow and outflow detection, annual revenue estimation, seasonal forecasting, and anomaly detection. The position also turns data from sources such as bank transactions and storefront sales into a defensible view of a business’s financial health, and sets technical standards for machine learning in Credit.

Requirements

At least 7 years of experience building machine-learning models.
Experience implementing models that shaped automated decisions involving money.
Production-grade Python skills.
Depth beyond classification models, such as time series and forecasting, anomaly detection, and natural language processing.
Proficiency with deep learning, transformers, and large language models.
Strong SQL skills and comfort working in a data warehouse.
Ability to explain models to credit officers, model validators, product managers, and engineers.
Ability to set modeling standards for data-science colleagues.
A quantitative degree or equivalent preparation.

Nice To Haves

Direct experience modeling bank transaction data.
Experience in small business lending or business credit cards.
Experience with revenue-based financing, merchant cash advances, or similar products.

Responsibilities

Build and own models for the lending program, including transaction classification, recurring inflow/outflow detection, annual revenue estimation, seasonal forecasting, and anomaly detection.
Turn messy data streams, including bank transactions and storefront sales, into a defensible view of a business’s health.
Deploy models into production with observability, monitoring, and tests.
Partner with Credit on underwriting policy, including data requests, follow-up information, and quantifying uncertainty.
Partner with Product on product design, including principal sizes, pricing, and repayment terms.
Assess where predictive modeling can meaningfully improve the underwriting program.
Set the technical bar for machine learning in Credit.