Work · Kaleidofin
From custom models to a multi-product lending platform
In about 18 months, per-client scoring models and a back-office app became a four-product risk platform. Two in three partners now use more than one product, partner onboarding is down by more than 80%, and platform ARR is up more than 200%.
Context
When I joined in May 2025, Kaleidofin deployed custom ki score models client by client, plus ki credit, a back-office app for co-lending and business-correspondent origination. Every new client meant bespoke work. ki credit assumed partners had their own loan origination system to call its APIs. Many smaller partners had none, and their policies varied too much to configure, so onboarding one took 6 to 12 months.
The hard call
Stop custom builds. Custom models were how Kaleidofin won deals, and saying no to them put near-term revenue at risk. I moved the company to standard, configurable products: one scoring framework, one origination app, one monitoring layer. We risked losing clients who wanted something bespoke. We gained a platform that onboards a partner in weeks and sells more than one product to each.
What I did
- Productised ki score: one model framework, 10 risk tiers, real-time APIs, no-code rules and five lender outputs. It is trained on more than five years of loan data, through Covid.
- Built Ki Originate, a configurable Android onboarding app run from an admin panel, in place of rigid APIs small partners couldn’t integrate.
- Used ki catalyst to risk-review new partners before onboarding them.
- Built ki monitor for life-of-loan monitoring, so score, monitor and catalyst share one loan-level risk spine.
- Rebuilt pricing so sales knew exactly what to sell and at what price.
My role
I lead product across the suite: strategy, roadmap and pricing, with a team of seven PMs, a designer and a product marketer. I report to the CTO and co-founder and run one backlog against one capacity pool across two group companies with separate P&Ls. Since August 2026 I also own marketing.
Results
| Measure | Before | Now |
|---|---|---|
| Platform products with paying customers | Custom models + back-office app | 4 |
| Partners on more than one product | Client-by-client deployments | Two in three |
| Partner onboarding | 6 to 12 months | ~1 month |
| Platform ARR | Baseline, May 2025 | Up more than 200% |
| KCPL lending partners | Many couldn't be onboarded | A new segment, opened by Ki Originate |
ki score delivers 20 to 30% more approvals with PAR90 under 3.5%. Company-wide, Kaleidofin reports more than 13 million customers and $11 billion of loans underwritten on ki score.
What I take from it
Custom work wins the first deal and caps the next hundred. Find what every client needs, make it configurable, and say no to the rest.
Next case: iCUE: putting LLMs into production for US lenders, starting where it was safe