Product Data Scientist

Selectfi•Buffalo, NY

5d•$130,000 - $170,000

About The Position

SelectFI is hiring a Product Data Scientist to identify valuable business problems and build data products, predictive models, pipelines, and application features to solve them. The role works directly with dealerships, lenders, sales, and customer success, and owns the work from understanding user needs and setting priorities through implementation and measuring results. The position includes end-to-end ownership of SelectFI’s lender approval models, lender data, and pipelines, using AI tools to prototype and build quickly.

Requirements

At least 4 years of experience building and deploying predictive models in production, ideally classification models on tabular data.
Strong Python skills for data work and modeling, including standard machine-learning libraries for tabular data.
Advanced SQL skills and hands-on experience with a cloud data warehouse; SelectFI uses Redshift on AWS.
Experience building and maintaining ETL pipelines, rather than working only from clean tables.
Experience shipping code through Git and pull requests.
Daily use of AI tools to reshape data, build pipelines, and create working features in unfamiliar front-end or back-end code, with validation of AI-generated work.
Comfort working directly with sales, customer success, and dealership staff and explaining data clearly to nontechnical audiences.
Ability to learn a business, identify priorities, define work, write tickets, and unblock oneself.
Inventiveness, curiosity, and a product mindset, with a focus on results for dealers and the business.
Bachelor’s degree in a quantitative field or equivalent practical experience.
Applicants must be authorized to work in the United States; the company does not currently offer visa sponsorship.

Nice To Haves

Startup experience.
Experience in consumer lending, credit risk, or auto or RV finance.
Experience in product management, solutions engineering, or another customer-facing role.
Hands-on web or app development experience, front end or back end.
Familiarity with dealership systems such as RouteOne, Dealertrack, or dealer CRMs.
Experience with model monitoring, feedback loops, or fair-lending review for credit models.

Responsibilities

Spend time with sales, customer success, and dealerships to learn their workflows and identify problems.
Identify valuable new data to collect and new uses for existing data, including product lines, dealer reports, and sales tools.
Define and prioritize work, create tickets, size potential impact, and keep plans visible to the team.
Rapidly prototype, often using AI tools, put results in front of users, and measure outcomes.
Build, retrain, and monitor models that predict lender approval and rates, and develop models for new product lines.
Report on model prediction performance, identify where and why models miss, and analyze dealer feedback to improve models and products.
Onboard new lenders by converting lender guidelines and rate sheets into rate tables and rules for dealership go-live dates.
Develop prediction strategies for new lenders when historical decision data is limited or unavailable.
Maintain lender eligibility rules, rate tables, and credit logic such as LTV, DTI, and score bands in the rules engine.
Build ETL pipelines from dealer systems, CRMs, credit bureaus, and data providers into the Redshift warehouse, and own data quality from source through model.
Use AI tools to extract information from PDFs and messy exports, scaffold pipelines, and write and test SQL.
Build initial working versions of app fields, screens, and reports when data or model changes require them, with the app team reviewing the work.
Adapt models to new verticals such as RV lending.
Answer business questions with data, including profitability analyses, funded loan-term analyses, and dashboards for internal teams and customers.
Work with sales, customer success, and dealership staff to obtain data, understand lender behavior, and investigate prediction errors.
During the first 90 days, learn the prediction pipeline, warehouse, and lender rules engine; ship pull requests and a significant model improvement or feature; extend a pipeline; participate in model and feature-priority reviews; and take an independently identified idea through to a shipped result.