The AI-native loan origination system.
A loan origination system that develops policies as well as runs them. Built by operators with two decades of consumer credit experience at PayPal and Varo.
Built on two decades of consumer credit.
building consumer credit at PayPal and Varo
on two consumer credit platforms
underwriting, policy, product, P&L
We're building Zendeci because we've spent two decades solving these problems the hard way.
Loan origination systems were built when policies didn't change.
So they're great at running policies — and terrible at developing them. When the market shifts, a risk team spends weeks pulling data, weeks building features, and weeks asking engineering to translate the new rule into the decisioning system. By the time the policy is live, the cohort that triggered it is already on book.
Zendeci is the first loan origination system where developing the policy and running the policy are the same product. Real-time decisioning sits on top of a workspace built for the people who change the rules.
Today: manual work spread across Excel, Tableau, notebooks, and the analyst's head. No dedicated tool exists.
Where today's loan origination systems live. They handle this layer well — but only this layer.
What happens when you want to tighten one rule.
A single threshold change. Today vs. with Zendeci.
Discover the signal
Analyst notices a DQ1 spike in a Tableau dashboard on Monday. Pulls last quarter's data into Excel. Spends two days isolating which segment is driving it.
Zendeci surfaces the DQ1 spike Monday morning with the segment already isolated. The recommendation is in the analyst's inbox: Tighten DTI cap to 43% on subprime auto · +$2.1M annualized · Confidence 0.84.
Build the new feature
Hand off to a data engineer to write the SQL. Two weeks of back-and-forth on definitions. Backfill, validate, document.
Analyst describes the feature in plain English. Zendeci generates the SQL, runs the backfill, versions it, and validates information value. Five minutes.
Ship it to production
Hand off to engineering to code the rule into the decisioning system. Two more weeks for translation, testing, and deployment. By the time it's live, the cohort that triggered it is already on book.
One click activates the rule in Zendeci's runtime. The audit trail flows through.
Six weeks to one afternoon. Same regulator. Same audit trail.
What we're building.
Four capabilities, one runtime. Available to design partners starting Q3 2026.
AI Credit Risk Manager
Surfaces policy recommendations from portfolio behavior. Quantifies the lift before you ship.
Data Connector
Wires into bureau data, Plaid cash flow, and alternative sources with versioned, governable pipelines.
Feature Store & Model Builder
Build features and train models from natural-language descriptions. Backfills, validates, and versions — production-ready, governance-built-in.
AI Decision Engine
The runtime. Real-time decisioning with reason codes, audit trail, and override paths built in.
The work, end to end.
Three steps of the credit lifecycle, rebuilt as one workspace.
From a thousand spreadsheets to one workspace.
Risk teams today operate from Excel, BI tools, bureau data feeds, cash flow data, and a separate loan origination system — passing files between analysts who each rebuild the same logic in a different syntax.
Zendeci replaces the stitching. Policy, features, and execution will live in one workspace, versioned and reviewable — designed to compress cycle time from months to days.
An audit-grade runtime.
Every decision Zendeci makes will carry its reason codes, feature values, and policy version with it. Built to support ECOA adverse action requirements, model risk governance, and the kind of audit trail your regulator already expects.
Re-price your portfolio overnight, not next quarter.
Designed to recompute PD continuously and route breach alerts to the desk that owns the relationship, with the proposed pricing delta attached.
What used to be a quarterly committee becomes an overnight run — built to compress manual work and free up the team.
We're working with a small group of lenders to shape the product.
Design partners get early access to Zendeci, weekly working sessions with the founding team, and direct input on the roadmap — in exchange for honest feedback and a pilot commitment when we hit GA. We're looking for two more lenders in 2026.
- US consumer credit lender
- Risk team of 3 or more
- Willing to commit 4 hours/month for 6 months
- Interested in early access pricing
- Mortgage-only lenders
- Pre-launch lenders
- No internal risk function
Be one of the first to use it.
Apply to be a design partner. We'll be in touch within one business day.