By Ayo Onikoyi

In early 2020, while working as a Software Engineer at Enyata, Suliat Oyekola joined a bold project to reinvent how creditworthiness is evaluated in the U.S. alternative lending space. The client? Kafene, a fast-growing fintech startup focused on helping underserved consumers access flexible financing.

At the time, Kafene was relying on a third-party product called Semnatic to make underwriting decisions. But it quickly became clear that this off-the-shelf tool wasn’t delivering the accuracy or agility Kafene needed to grow. So they brought in Enyata — and Suliat stepped up to build something better from scratch.

Over the next six months, she built a new AI-powered underwriting system that would eventually replace Semnatic completely. The result: a 60% improvement in underwriting accuracy and a stronger, smarter way to evaluate customers who were often overlooked by traditional credit scoring methods.

Building the Brain Behind the Credit Decisions

Suliat’s work began with building what she calls the “Underwriting Knowledge Base” — a smart system that combines data from major credit bureaus like Experian, Clarity, and Giact, and evaluates it using machine learning models.

“We weren’t just pulling in data — we were teaching the system how to understand it,” she explains. “We trained models on real historical data, so the system could spot patterns in how people behave financially, not just what their credit scores said.”

This was especially important for Kafene’s mission. Many of their customers don’t have traditional credit histories, so Suliat’s system had to go beyond the surface — analyzing everything from income patterns to identity verification data, and making fast, real-time decisions on loan approvals.

The AI Method That Changed the Game

To power this innovation, Suliat developed a custom AI technique internally dubbed the “Weighted Bureau Fusion Model.” The idea was simple but powerful: instead of treating each credit bureau as a single source of truth, her model gave each one a different weight depending on the type of customer and the quality of the data returned.

“It was like giving each bureau a vote — but some votes counted more than others,” she says. “That helped us avoid misleading signals and make decisions that were closer to how a human expert would think — but much faster.”

This method dramatically reduced false positives and allowed Kafene to say “yes” to more qualified applicants — without taking on unnecessary risk.

Data That Learns, Decisions That Improve

Beyond just building the system, Suliat also implemented a data monitoring and feedback loop. She defined key performance metrics — like approval rates, default rates, and model drift — and set up dashboards that allowed Kafene’s product and risk teams to continuously improve the models based on real-world outcomes.

The impact was immediate. More people got approved. Fewer loans went bad. And Kafene gained a reputation for being a tech-forward lender that truly understood its customers.

From Lagos to New York: A Story of Tech for Good

For Suliat, the project wasn’t just about writing code. It was about using technology to create real-world impact.

“I’ve always been passionate about building tools that solve real problems,” she says. “With this project, we weren’t just helping a business grow — we were helping everyday people get access to the financial tools they need to live better lives.”

Today, Kafene continues to scale, and the underwriting system Suliat helped build remains at the core of its decision-making engine.

Her journey — from developing simple programs in Nigeria to building AI systems powering U.S. fintechs — is a reminder of the global reach of talent and the power of software to open doors for people everywhere.