The core engine behind every Karus decision
Karus applies predictive technology across four layers: where the loans come from, how the model underwrites them, how the engine prices them, and how it all connects to your systems. Here is what each layer does.
Loan sources
Access Layer
Structural preference and first-look routing into quality loan flow.
L1 + L2 model
The industry dataset, with each customer's own data and strategy layered on top.
Decisioning
Trajectory pricing and loss timing curves
Pricing
Loan level win rate curves ensure accurate pricing for dealer and buyer
Connection layer
Decisions flow through API into the systems you already run.
.svg)
.svg)
.svg)
.svg)
Delivered to
Originator's pipeline
The decision and price delivered to the lender who books the loan.
Investor's ledger
The same underwriting standard delivered to the buyer who funds it.
It starts with access, not the model
Most platforms start with a model. Karus starts with access. The best auto loans never reach the open market: dealers route them to preferred lenders first, and a platform without a position in that flow underwrites whatever is left.
Karus holds structural preference inside the dealer routing channels themselves. Through partnerships with Automatic USA, CRIF, DOFC, and Credit Smarts and through direct relationships with related finance companies and credit union programs, Karus sees quality loan flow before it reaches the broader market. That first-look position is the foundation under everything that follows.
A competitor can build a model. Replicating this access is far harder, because it is earned through years of integrations and relationships rather than bought. Access without an accurate model wastes the loans. A model without access never sees the best flow. Karus is built to do both.
Two layers: the industry model and your
customization
Once a loan reaches the platform, two layers of data decide how it is underwritten.
Layer 1: the industry model
Layer 1 is the base model, trained on the most complete auto loan dataset in the industry: nationwide, multi-decade payment history across millions of loans. That depth is what lets Karus read a borrowers credit trajectory and project a full loss curve rather than a loss average. It is also what separates a credible projection from a curve fit to too few outcomes.
.svg)
Layer 2: your customization
Layer 2 is your own data and strategy layered on top. It turns the base model into a scorecard unique to you, tuned to your yield target, loss tolerance, dealer mix, and credit appetite. Two customers running the same base model reach materially different decisions, because each has programmed the engine to its own strategy.
Layer 1 brings the industry's data. Layer 2 makes the model yours.
Decisioning and pricing, loan by loan
With access and the model in place, the engine makes the decision and sets the price on every loan.
01
Trajectory pricing
Karus prices each loan to the borrower's credit trajectory, not a static score. A 650 trending toward 700 is priced to capture yield; a 650 trending toward 600 is priced to protect it. The rate reflects the loan's actual risk at acquisition.
02
Loss timing curves
The engine models when losses hit, not just how large they get. Roll-to-default behavior and recovery rates shape a full loss curve for every vintage, so the platform can catch a bad loan before approval rather than after it surfaces in performance.
03
Loan-level pricing in seconds
Every loan is decisioned and priced individually, against the borrower, the vehicle, and the dealer's expectations, in real time. That is what lets a lender win dealer acceptance without giving away yield elsewhere in the waterfall.
04
Programmable strategy
None of this is fixed. You program the engine to the outcomes you commit to deliver, and adjust as rates, loss curves, or funding costs move, rather than waiting for a new model build.
Built to fit your stack, not replace it
Karus runs inside the workflow you already have. The platform connects through API to major loan origination and servicing systems like Dealertrack and RouteOne, so decisions and prices flow into an originator's pipeline in real time, and into a structured buyer's ledger, without rebuilding either one. You go live alongside your current process, with no cutover downtime and no interruption to the dealers already sending you business.
You choose where the engine sits in the decision flow: fully automated for the tiers you trust, with a human in the loop wherever you want one. Standardization and a complete audit trail come built in, which is what capital partners need to underwrite the book.
Loan origination systems
Servicing platforms
Dealer routing channels
AI underwriting. Programmable credit strategy for auto finance.
Karus is not competing in the AI underwriting category. It is defining a new one: programmable credit strategy for auto finance. The result is portfolio economics built on loan access, accurate loan-level pricing, and consistent outcomes that competitors cannot easily replicate.
Generic AI underwriting
Borrower score
Programmable credit strategy
Programmable engine
See your portfolio run through Karus
Bring us a sample of your origination volume. No commitment to integrate
— just a side-by-side of your manual decisions and ours.
Bring us a sample of your origination volume. We will run it through Karus and show you, loan by loan, where the platform would have approved differently, priced differently, or held back.
Send us a portfolio file. We will rebuild it loan by loan through Karus and show you the projected net unlevered yield, the loss timing curves, and where the selection differs from manual.