Underwrite.ai

Overview

Underwrite.ai sells one thing: a custom credit risk model, delivered as a decision through an interface. There is no origination workflow, no document pipeline and no data business around it. A lender supplies anonymized historical loan and application records, the company builds and maintains a model calibrated to that specific portfolio and lending environment, and the resulting decision returns in two to three milliseconds. What distinguishes the record is how much of the method is published. The technique is named as gradient boosting rather than described as artificial intelligence in the abstract, and the case for it is made concretely: traditional scorecards use logistic regression over fifteen to twenty variables and assume linear relationships, while a boosted model works across hundreds and finds threshold effects and interactions no analyst encoded. Explainability is addressed by naming the method, Shapley additive explanations, and stating that every decision carries documentation of which factors drove it, with worked examples of contribution decomposition. Fair lending is treated as a product capability rather than a compliance assertion. The company publishes disparate impact analysis tools, states that its models are designed to avoid variables that proxy for protected classes, explains the proxy mechanism using postal code and historical housing segregation, and generates reports comparing outcomes across demographic groups so a lender can find bias before deployment rather than after. Published results are unusually specific. An online installment lender with a 32.8 percent first payment default rate and overall defaults above 60 percent saw early default fall to 8.5 percent using the model as its sole underwriting method. A Korean market study reports a well tuned logistic regression at an area under curve of 0.906 against 0.958 for the boosted model. A United Kingdom auto lender moved from two to three days to two to three milliseconds. The company has also modeled populations with little or no bureau data, including unbanked borrowers in rural Mexico and less developed Philippine provinces. Founded 2015 and based in Boston, operating as SVM Ventures LLC Series Underwrite.ai, with a published price, a public pricing page and a 30 day free trial.

Details

Similar Tools

• AIZEN Global Credit & Lending
• Aloan Credit & Lending
• Barkr Credit & Lending
• Bizbaz Credit & Lending