Ex-Uber Engineers Raise $3M for Fintech Fraud Risk Startup Sperta

The Genesis of Sperta: From Uber's Risk Team to a Rules Engine as a Service

While collaborating within Uber’s risk division, Yifu Diao and Ming Fang identified escalating fraud challenges accompanying the company’s rapid expansion. Diao’s initial assignment, commencing in 2014, involved the development of Mastermind, a rules engine designed to combat fraudulent activities.

Mastermind's Evolution and Impact

This engine, according to Diao, empowered risk analysts to implement and deploy rules independently, without requiring engineering support. Initially conceived for fraud detection, its applications broadened to encompass safety protocols and customer service operations.

By the time of Diao’s departure, Mastermind was processing an impressive 10,000 decisions every second. When Fang joined the Uber risk team in 2016, the need arose to integrate Mastermind into the dispatch process – the point at which drivers and riders are connected.

Scaling Mastermind for Critical Systems

Dispatch represented a particularly crucial system for Uber, demanding significantly greater scalability and stricter latency constraints than previous applications. Fang undertook substantial optimizations to Mastermind, enabling this new functionality and substantially reducing instances of fraud, as Diao recalls.

Currently, Mastermind is utilized by hundreds of analysts and operations personnel, facilitating real-time decision-making across numerous user interactions.

Identifying a Broader Need for Rules Engines

Diao’s final year at Uber was dedicated to a lending product, which also leveraged Mastermind for underwriting purposes. Subsequently, joining a credit card startup, he recognized a similar requirement for a rules engine to effectively manage fraud, credit, and compliance risks.

“Explainability is paramount for fintech companies, enabling them to demonstrate decision-making processes to regulatory bodies,” Diao explained. This realization underscored the demand for a rules engine offered as a service.

The Founding of Sperta

Diao immediately considered Fang, who was then leading Google’s Cloud AI Feature Store. Having previously discussed potential startup ventures, the pair decided to collaborate and establish Sperta this June. The name, meaning “expert” in Italian, reflects the company’s core mission.

Sperta's Mission and Differentiation

Sperta aims to assist financial services and technology companies in automating decisions and mitigating fraud, credit, and compliance risks. The founders believe that many existing rules engine offerings, with their emphasis on no-code UIs, will prove inadequate when dealing with complex logic.

According to Diao, Sperta’s primary competitive advantage lies in its expression language. Targeting analysts and data scientists already proficient in SQL, they opted to create a language with similar syntax. This approach proved successful with Mastermind, allowing analysts to become proficient within just one week.

The Importance of Rigorous Testing

Diao emphasizes that rules are ineffective without thorough testing. High false positive rates can negatively impact growth, potentially leading to customer attrition for financial institutions and businesses.

“We are acutely aware of this,” Diao stated. “Therefore, we are incorporating rule unit tests for verification, backtesting for performance evaluation, and percentage rollouts for the safe implementation of rule changes.”

Building a Comprehensive Risk Decision Platform

To address the complete needs of its customers, Sperta is developing a risk decision platform centered around its rules engine. The platform integrates with data vendors and enables analysts to transform features obtained from these sources.

Customers can also integrate their own models, benefiting from Sperta’s streamlined interface. When automated decisions are not feasible, cases can be directed to Sperta’s case management tool.

“While models provide probabilistic predictions, rules deliver explainable decisions and deterministic actions,” Diao noted. “We are enthusiastic about empowering the internet to make faster, more informed decisions.”

Seed Funding and Future Development

Just two months after its inception, Sperta secured $3 million in seed funding, co-led by Kindred Ventures and Uncork Capital, with contributions from several angel investors. The company is now publicly disclosing this funding round.

Fang highlighted that Sperta simplifies the integration of models for its customers. “We can structure the data they need to detect risk and make decisions,” he explained to TechCrunch. “We can structure the way they use data to ensure their decision-making is secure and efficient.”

Targeting Fintech and Financial Institutions

A rules engine is particularly crucial for fintech companies during the onboarding process. Fraud prevention is a top priority for all financial institutions, both traditional and fintech, and the increasing volume of online transactions elevates the risk of fraud.

Consequently, fintechs and financial institutions represent key target customers for Sperta.

Diao indicated that competitors typically charge around $1 per decision, and Sperta intends to offer a more competitive pricing structure.

The company plans to allocate its new capital primarily to hiring and aims to have a Minimum Viable Product (MVP) available by the end of the year.

Investor Confidence in Sperta's Potential

Kanyi Maqubela of Kindred Ventures believes that Diao and Fang’s work at Uber was “groundbreaking” in the field of fraud prevention. “A generalized, powerful version of a decision engine is sorely needed in today’s market,” he stated. “Software companies need a solution that is sufficiently sensitive to allow both customization, yet maintain compliance.”

Andy McLoughlin, managing partner at Uncork Capital, agrees that the founders’ experience building a rules engine at Uber provides them with a significant advantage. “In our diligence, we heard repeated dissatisfaction with existing solutions lacking some of the advanced features Sperta delivers in v1.0,” he noted. “The competitors have primed the market, but we see a huge opportunity to deliver the correct solution.”