Tag: Fraud

One of America’s most successful women entrepreneurs is at it again, striking out this time in partnership with her husband and giving herself the CEO title for the first time. Neha Narkhede, a software engineer who cofounded data-streaming software firm Confluent in 2014 and served as its chief technology and product officer for more than five years, has a new company coming out of stealth Thursday, she shared exclusively with Forbes.

Narkhede, 38, and her husband Sachin Kulkarni, 39, a former engineering executive at Meta Platforms, founded Oscilar in 2021 to decrease the risks involved with online transactions—mainly the risk of loan defaults and fraud with things like financial transactions and insurance—using artificial intelligence. They are funding it themselves with $20 million, have hired about two dozen employees and say they have signed up dozens of customers, many of which are fintech firms with an average of 500 employees. Narkhede is the chief executive officer and Kulkarni is chief technology officer of the remote-work-first firm, which has a small office in Palo Alto, California.

Narkhede says the goal with Oscilar is “making the internet safer.” To do that, she and Kulkarni spoke to 100 risk experts at fintech and other companies and learned that existing risk models rely on incomplete, outdated information about user behavior and do not always use the most updated machine learning techniques. They believe they can protect online transactions from fraud and theft more quickly and accurately with less engineering support than others by combining well-sourced data with AI. The problem is real and growing. U.S. consumers reported losing $8.8 billion to fraud in 2022, an increase of nearly $2.6 billion over 2021, according to the Federal Trade Commission.

“The key benefit is that we remove the need for a company to use engineers for risk assessments since no coding is required,” Narkhede says. “Businesses decide ahead of time what data they want to be analyzing, and we set up the programming to ensure our AI technology brings in the data to advise on the risk of every transaction, leaving it to the risk analysts so they run tests and approve tweaks to the model.”

Narkhede is a born-and-bred engineer. After earning her bachelor’s in computer science from Savitribai Phule Pune University in India, she immigrated to the United States for a master’s degree in computer science from Georgia Tech in Atlanta, from which she graduated in 2007. She then had stints as a software engineer at Oracle and LinkedIn.

At LinkedIn, she and two colleagues—Jay Kreps and June Rao—co-created open source messaging system Apache Kafka to handle the professional networking site’s huge amount of incoming data. In 2014, the Apache Kafka founders left LinkedIn to found Confluent, which helps organizations process large amounts of data using Apache Kafka. By early 2019, Confluent had raised more than $200 million from venture capital firms; that helped Narkhede land on Forbes’ annual list of America’s Richest Self-Made Women in 2019 (and stay on since then). Confluent went public in June 2021, spiked to a $24.75 billion market capitalization and made Narkhede a billionaire for several months, before the stock fell 77%. She sold close to $170 million worth of Confluent stock (before taxes) prior to and after Confluent’s 2021 IPO. Forbes estimates that she’s currently worth about $475 million.

In January 2020, she stepped down as Confluent CTO but kept her board seat. Back then, the idea for Oscilar had already taken root. “When I was involved in Confluent day-to-day, I saw companies that use Apache Kafka struggle with building their fraud and [credit] risk decisioning systems,” Narkhede says. “That’s when the seed was planted in my brain.”

The problem Oscilar is meant to address, Narkhede describes, is that existing risk detection systems have a hard time pulling together disparate data sources, can be slow to adapt to new input and can be hard to customize. Oscilar’s product—a constantly-training, no-code model that users can, alongside Oscilar’s team, customize and fine-tune—attempts to address the gaps Narkhede sees in existing systems.

Notably, the company received offers of venture capital cash—“all inbound interest,” Narkhede says—but she turned them down in favor of bootstrapping. “Self-funding has provided us with the autonomy to move fast,” Narkhede says, adding she will likely be open to outside funding in the far-out future. She says the initial funding should provide Oscilar with “several years of runway.”

Narkhede and Kulkarni’s 9-to-5 these days is 9 p.m. to 5 a.m.—their sleeping hours. To manage a team with employees spanning North America and Europe on top of their two-year-old son, the couple tucks in for the night at 9 p.m. and begins their work days at 5 a.m., Kulkarni says. They’ve done so ever since they started Oscilar—around the same time their son was born.

To run Oscilar, Narkhede and Kulkarni made sure they have a clear separation of responsibilities, advice they received after consulting other cofounder couples, according to Kulkarni. Even though both have highly technical backgrounds, he is responsible for the engineering side of the company, while Narkhede spends more time on the operations and clients.

Oscilar has plenty of competition from companies like DataVisor, Provenir, Sift and Alloy, as well as bigger outfits like Google that have the bandwidth to build their own risk models.

Like most risk-detecting models, Oscilar uses a combination of customer biographical information, customer transaction history and third-party data from credit bureaus and other sources. Oscilar also touts something called a “semi-supervised” machine learning algorithm, which means it combines labeled data, which includes an “outcome” like credit risk score, and unlabeled data—which does not, and is therefore easier to process—into one model. That approach, too, is not necessarily unique.

But Oscilar’s team, which customer Henry Shi, chief operating officer at $1 billion (sales) fintech firm Super, describes as super responsive and easy to work with, is a large part of what makes the company and its product stand out, says Shi. Oscilar’s 25-person team has members with data science or engineering experience—often specifically in building AI risk models—at places like Google, Uber, Meta and Confluent.

The “secret sauce,” as Narkhede puts it, is the team’s technical ability to gather high-quality data, create a comprehensive understanding of a user’s behavior, and build quickly and automatically updating AI models. Narkhede says she uses the terms AI and ML—for machine learning— interchangeably.

As Oscilar looks toward its first weeks out of stealth, the company is actively hiring and aims to further expand its customer base to fintech companies of different sizes and across different sectors—using its product to lead the way despite its CEO’s big reputation.

Super COO Shi, for example, who met Narkhede at a fintech conference in fall 2022 and became a customer soon afterward, was first drawn in by Oscilar’s product because it was low-code, customizable and dynamic, he says. “I didn’t even realize she was the cofounder of Confluent until the very end of our conversation.”

This article was updated on March 30 to clarify that Uber was a customer, not Lyft.