The Log and the Ladder
In the summer of 2011, inside LinkedIn's Mountain View campus, a software engineer named Neha Narkhede was staring at a problem that had no name. LinkedIn was growing at a rate that overwhelmed every system the company had built to capture what its users were doing — every click, every profile update, every connection request — and the torrent of data was breaking things. The standard approaches to data integration, the batch-oriented extract-transform-load pipelines that had served enterprise computing for decades, were fundamentally inadequate for a social network generating billions of events daily. You could store data at rest or you could process it in transit, but you could not do both simultaneously, and doing neither well meant that LinkedIn's internal systems were drowning in latency. Narkhede, along with her colleagues Jay Kreps and Jun Rao, began building something that did not yet exist: a distributed streaming platform that treated data not as a thing to be warehoused but as a continuous, replayable flow. They called it Kafka — after the writer, because the system was, in Kreps's telling, "optimized for writing." The technical innovation was a distributed commit log that combined the persistence of a database with the real-time throughput of a messaging queue, and it would become one of the most consequential pieces of open-source infrastructure of the 2010s, eventually processing more than seven trillion messages daily across more than 80 percent of the Fortune 500.
But this is not, fundamentally, a story about a commit log. It is a story about a woman from Pune, India, who received her first computer at age eight, read biographies of Indira Gandhi and Kiran Bedi at her father's insistence, earned a master's degree at Georgia Tech, and then — through a combination of deep technical fluency, entrepreneurial audacity, and a refusal to wait for permission — built and co-founded two companies, accumulated a personal fortune exceeding half a billion dollars, and did it all before turning forty. The log was the vehicle. The ladder was hers to construct.
By the Numbers
The Narkhede Record
$9.1BConfluent IPO valuation, June 2021
~6%Narkhede's ownership stake in Confluent at IPO
$520MEstimated net worth (Forbes, 2023)
80%+Fortune 500 companies using Apache Kafka
7T+Messages processed daily via Kafka globally
$20MSelf-funded investment to launch Oscilar
60+Oscilar employees within first year of emerging from stealth
A Computer in Pune
Pune in the early 1990s was not yet the IT boomtown it would become. It was a city of universities, defense establishments, and the quiet middle-class ambitions of engineering families — the kind of place where a mechanical engineer father could reasonably expect his daughter to study hard, marry well, and perhaps land a government posting. Neha Narkhede's father had different instincts. When she was around eight years old, he bought the family a personal computer — an unusual acquisition for an Indian middle-class household of that era — and watched his daughter become consumed by it. She began teaching herself to code with the limited resources available, and her parents, rather than directing her curiosity toward more conventional pursuits, fed it. Her father, in particular, curated a private curriculum of female ambition. He told her stories of Indira Gandhi, India's first woman prime minister. He spoke of
Indra Nooyi, the Madras-born executive who would eventually run PepsiCo. He gave her books about Kiran Bedi, the first woman to join the Indian Police Service. "He picked examples from many different walks of life," Narkhede later told CNBC. "This cultivated a sense of empowerment in me, and made me believe that if people like me can do this impossible thing, then I can too."
There is something almost too neat about the origin story — the prescient father, the precocious daughter, the computer as totem — but its significance lies less in its narrative arc than in what it reveals about the specific cultural machinery that produced Narkhede. She was not a rebel against her upbringing but its apotheosis. Her parents instilled in her what she would later call a "growth mindset," the belief that anything could be learned given sufficient time and effort. They taught her to value education as the primary instrument of social mobility. And they ensured, crucially, that she encountered women who had already accomplished the impossible, so that impossibility itself became a solvable problem rather than an immutable condition.
She studied computer science at the Pune Institute of Computer Technology, part of Savitribai Phule Pune University, and in 2006 — at age twenty-one or twenty-two — she left India for the Georgia Institute of Technology to pursue a master's in computer science. The decision to emigrate was deliberate: she wanted a career in technology, and the center of gravity for that career was in the United States. Georgia Tech was rigorous, practical, and connected to the Silicon Valley ecosystem in ways that mattered. She completed her degree in 2007 and stepped directly into the American technology industry.
Oracle, LinkedIn, and the Shape of Data
Narkhede's first job was at Oracle, where she spent two years as a lead engineer working on hierarchical faceted search in Oracle Text. It was technically demanding work inside one of the most established database companies on Earth, and it gave her a foundation in how enterprises stored, indexed, and retrieved information. But Oracle was a company oriented toward data at rest — structured, queryable, patient. Narkhede wanted something faster.
In February 2010, she joined LinkedIn as a software engineer. LinkedIn at that time was in the midst of explosive growth, transforming from a professional rolodex into a social network with hundreds of millions of users, and its data infrastructure was struggling to keep pace. Every user action — a profile view, a search query, a connection request, an ad impression — generated data that needed to be captured, routed, and processed in something approaching real time. The existing systems were a patchwork of point-to-point integrations: data flowed from application A to database B through a bespoke pipeline, and from database B to analytics system C through another. The result was a tangled web of dependencies, each one a potential point of failure, and none of them fast enough.
Jay Kreps, who would become Narkhede's co-founder, was a principal staff engineer at LinkedIn. He had been thinking about the problem at the level of first principles: what if, instead of building a separate pipeline for every data movement, you built a single, central platform that captured every event as it happened and made it available to any system that needed it? Jun Rao — a database researcher with deep experience in distributed systems who had previously worked at IBM's T.J. Watson Research Center — brought the rigor of formal systems design. Narkhede brought an engineer's obsession with making things work at scale. The three of them, along with a broader team, began building what would become Apache Kafka.
The key insight was architectural: rather than treating messaging, storage, and processing as three separate concerns, Kafka unified them in a single abstraction — the distributed commit log. Messages were written to a log, persisted durably on disk (unlike most messaging systems, which discarded messages after delivery), and could be consumed by multiple subscribers at their own pace. This meant you could use Kafka for real-time messaging and for replaying historical data, which was a category-defying capability. As Narkhede would later explain in keynotes across the world, "Systems are giving up correctness for latency, and I'm arguing that stream processing systems have to be designed to allow the user to pick the tradeoffs that the application needs."
Kafka was open-sourced in 2011. By the time Narkhede left LinkedIn in September 2014, it had become the backbone of the company's entire data infrastructure, processing hundreds of billions of events per day, and its adoption was spreading across the industry. The decision to open-source was not altruistic — it was strategic, a way to build a community, establish a standard, and create the conditions for a commercial enterprise that could sit on top of the open-source project.
The Founding Arithmetic
On September 23, 2014, Narkhede, Kreps, and Rao incorporated Confluent. The founding was not a leap into the unknown so much as a calculated bet on a known commodity. Kafka was already widely adopted, already proven at LinkedIn scale, and already attracting attention from enterprises struggling with the same data-in-motion problems. The question was not whether there was demand but whether a startup could capture it before the hyperscalers — Amazon, Microsoft, Google — built their own managed Kafka services and commoditized the technology.
LinkedIn itself put in $500,000 in initial funding, a modest investment that functioned as both an endorsement and a bridge. Data Collective (DCVC) led a $3 million seed round, followed quickly by a $6.9 million Series A. The early capital was small by Silicon Valley standards, but the founding team had something most enterprise startups do not: a product that already had traction. "We weren't building from zero," Narkhede would later reflect. The zero-to-one phase of Confluent was unusual because the technology had already crossed that threshold; the company's job was to build the zero-to-one for the business.
You should start a company when you feel that launching it is the only vehicle to see the change you want in the world.
The co-founder dynamic was complementary in a way that reads, in retrospect, as almost engineered. Kreps — cerebral, vision-oriented, a systems thinker who had literally written the book on distributed logs — would serve as CEO, setting strategic direction and articulating the category-creation thesis: that "data in motion" was as fundamental as data at rest, and that Confluent would build the foundational platform for it. Rao, the distributed systems architect, would focus on the core technology. Narkhede, as CTO and later Chief Product Officer, owned the product and engineering organizations — the dual mandate of building the technology and ensuring it could be sold.
Early on, they made a series of decisions that would define the company's trajectory. First, they established Confluent as an enterprise company from the outset, targeting large organizations with complex data needs rather than pursuing a bottoms-up developer-adoption model alone. This was unusual for an open-source company; many started with a free community edition and hoped adoption would percolate upward. Confluent did both — investing heavily in developer evangelism to grow the Kafka community while simultaneously building a direct sales motion aimed at Fortune 500 accounts.
Second, they decided early on what would remain open source and what would be proprietary. The core Kafka engine stayed open. The enterprise features — security, multi-datacenter replication, management tooling, and eventually the fully managed cloud service — would be licensed. This was the open-core model, but Narkhede and Kreps approached it with unusual discipline, treating open source not as a business model but as a distribution strategy. "Open source isn't a business model, it's a distribution strategy," Narkhede would say, with the flat certainty of someone who had thought about the distinction for years.
Third, they bet on the cloud before it was obvious that enterprise infrastructure companies needed to. In 2017, Confluent launched Confluent Cloud, a fully managed Kafka service that removed the operational burden of running the platform. This would prove to be the single most important strategic decision in the company's history, transforming Confluent from a software vendor into a consumption-based cloud service with recurring revenue — and positioning it ahead of the managed Kafka offerings that Amazon and other cloud providers would eventually launch.
Running Two Companies Inside One
The challenge of building an open-source company is that you are, in effect, running two enterprises simultaneously. There is the community — the tens of thousands of developers contributing to, deploying, and evangelizing Kafka — which requires investment, attention, and a credible commitment to the open-source ethos. And there is the commercial business — the sales teams, the enterprise features, the cloud service — which must generate revenue sufficient to justify venture-capital valuations. The tension between these two imperatives is not theoretical. It manifests in specific decisions: which features go into the open-source release versus the commercial product? How much engineering time is allocated to community contributions versus proprietary development? When does developer evangelism end and sales begin?
Narkhede navigated this tension by building a developer-evangelism function that served both purposes. Confluent invested heavily in content — blog posts, conference talks, technical documentation, a Kafka Summit conference series — that educated the market about event streaming as a paradigm. This was category creation in its purest form: before you could sell the product, you had to sell the idea that real-time data streaming was a fundamental capability every company needed. Narkhede was a central figure in this effort, delivering keynotes at conferences around the world — QCon, dotScale, Kafka Summit in San Francisco and London — that articulated the vision with both technical depth and strategic clarity.
Her 2016 QCon keynote, titled "ETL Is Dead; Long-Live Streams," was a landmark moment. In it, she argued that the entire paradigm of batch-oriented data movement — extract, transform, load — was obsolete, and that the future belonged to continuous, real-time event streams. It was a provocative claim, but it was grounded in LinkedIn's experience and in the growing evidence that companies like Goldman Sachs, Netflix, and Uber were adopting Kafka for mission-critical use cases.
The commercial business grew in parallel. Confluent's go-to-market evolved through phases that Narkhede would later describe with precision: founder-led sales first, where the co-founders themselves closed the earliest enterprise deals; then a transition to a structured sales organization, with the founders gradually withdrawing from individual accounts; and finally the addition of outbound sales, once the inbound pipeline was mature enough to demonstrate repeatable demand. The order mattered. Narkhede was emphatic that attempting outbound sales before achieving inbound pull was a mistake — it inverted the natural learning loop and led companies to push products the market had not yet asked for.
The IPO and the Question of Wealth
Confluent's funding trajectory tells the story of a company that was perpetually undervalued by the market until it wasn't. A $6.9 million Series A in 2014. A $24 million Series B in 2015. A $50 million Series C in 2017. A $125 million Series D in January 2019, at a $2.5 billion valuation. A $250 million Series E in April 2020, at $4.5 billion. Each round brought new investors — Sequoia Capital, Index Ventures, Benchmark — and each round reflected the market's growing recognition that data streaming was not a niche but a platform category.
On June 24, 2021, Confluent went public on the Nasdaq under the ticker CFLT, pricing at $36 per share and opening at $44, giving the company a first-day valuation of approximately $9.1 billion. By the end of its first day of trading, it had risen further, and the company raised $828 million in capital. Narkhede, who owned approximately 6 percent of the company, held shares worth hundreds of millions of dollars.
The wealth was staggering by any measure, and particularly by the measure of her origins. A woman born in Pune, India, into a middle-class family — daughter of a mechanical engineer and a homemaker — had co-created a technology used by the majority of the world's largest companies and built a personal fortune that would land her on Forbes' list of America's Richest Self-Made Women every year from 2020 onward. In 2023, she was ranked 50th, with an estimated net worth of $520 million. The Hurun India Rich List placed her among the wealthiest Indians in the world. She was, at thirty-seven, the youngest self-made woman entrepreneur on India's rich list.
But wealth in technology is volatile. Confluent's stock, like those of many high-growth SaaS companies, fluctuated wildly after its IPO. In 2022, the broader tech correction cut the company's market capitalization significantly, and Narkhede's paper wealth fell from a peak near $925 million to approximately $490 million. She was not, however, a passive observer. She had already stepped back from her day-to-day operational role.
The Departure and What It Reveals
In January 2020, Narkhede transitioned from her CTO role to become a board member of Confluent. The departure was not acrimonious, but it was consequential. She had spent nearly six years building the company's product and engineering organizations from a handful of people to a global operation, and she was leaving before the IPO — before the moment when founders typically cash in the culmination of their work. The move suggested that she was driven by something other than the desire to see a financial event through to its conclusion.
What followed was a period of deliberate exploration. She became an investor and advisor to a portfolio of startups — Material Security, Stytch, Supabase, Tabular, StarTree, Yugabyte, Abacus AI, Cortex Data, and others — accumulating exposure to the breadth of the infrastructure and security landscape. She was scouting, whether she knew it yet or not, for the next problem worth building a company around.
I learned to embrace my own unique background that truly makes me who I am today. The more you get comfortable with your own identity and see ways in which you are adding value because of that identity — that's the right way to go about this.
The answer arrived through Confluent itself. While serving as CTO and Chief Product Officer, Narkhede had watched the use cases flowing through the platform. She saw Kafka deployed for recommendation engines, for real-time pricing, for supply-chain coordination. But the fastest-growing and most urgent category of use case, the one that appeared again and again across industries, was fraud detection and risk decisioning. Banks, fintechs, e-commerce platforms — all were struggling with the same fundamental challenge: detecting fraudulent activity in real time, before the damage was done, using systems that could process thousands of signals in milliseconds.
The existing solutions were, in Narkhede's assessment, deeply inadequate. They were fragmented — a point solution for identity verification, another for transaction monitoring, a third for credit decisioning — and they operated in silos, unable to share risk signals across the customer lifecycle. They were slow to adapt, relying on static rules rather than machine-learning models that could evolve with the threat landscape. And they required armies of engineers to implement and maintain, which meant that only the largest institutions could afford comprehensive fraud protection.
Oscilar: The Second Act
In 2021, Narkhede co-founded Oscilar with her husband, Sachin Kulkarni. Kulkarni had spent over eleven years at Facebook (later Meta), where he had worked on infrastructure and data systems at enormous scale. The two of them invested $20 million of their own money — no venture capital, no outside investors — and began building in stealth.
The decision to self-fund was deliberate and revealing. Narkhede had raised hundreds of millions in venture capital for Confluent and understood both the benefits and the constraints of that model. For Oscilar, she wanted the freedom to iterate without the pressure of quarterly board meetings and growth-at-all-costs expectations. She also had the means: a half-billion-dollar net worth provides a remarkably effective substitute for institutional capital.
The founding team recruited from the engineering organizations of Google, Uber, Meta, Apple, and Confluent itself — a roster of talent that reflected the founders' combined networks across the highest echelons of Silicon Valley engineering. The company emerged from stealth in March 2023, when Forbes broke the story of its existence.
Oscilar's product is an AI-powered risk decisioning platform that unifies fraud detection, credit decisioning, compliance monitoring, and identity verification into a single system. The core thesis is that risk is a data problem, not a point-solution problem, and that the fragmented approach — buying one vendor for KYC, another for transaction monitoring, another for AML — creates the data silos that make fraud possible in the first place. If you unify the data, you can share risk signals across the entire customer lifecycle, from onboarding to transaction to ongoing compliance, and you can do it in real time.
The technical architecture reflects Narkhede's Kafka heritage: real-time data processing is not an add-on but the foundation. "Real time is in our DNA," she has said. The platform ingests and analyzes thousands of signals in milliseconds, using machine-learning models that adapt to evolving fraud patterns — synthetic identities, deepfakes, coordinated fraud networks — without requiring customers to write code. The no-code aspect is critical: Narkhede saw at Confluent that even companies willing to invest in fraud prevention were constrained by the engineering talent required to implement and maintain traditional solutions.
By 2024, Oscilar had surpassed sixty employees and was serving customers including SoFi and Trans Pecos Banks. The company was operating in North America, Europe, Latin America, and the Middle East. The results were, by the company's account, striking: a 90 percent reduction in fraud for some customers, a 40 percent reduction in AML operations costs for Trans Pecos Banks, and a 70 percent reduction in alert review time.
Married Co-Founders and the 5 A.M. Calendar
There is one biographical detail that most profiles of Narkhede mention only in passing but that deserves a longer look. She co-founded Oscilar with her husband. They are raising a young child together. They fund the company from their personal wealth. They work side by side, every day, on the same product.
The arrangement inverts every piece of conventional startup wisdom about co-founder relationships. The standard counsel is to avoid romantic partners as co-founders — the risk of contamination between personal and professional conflict is considered too high. Narkhede and Kulkarni have addressed this not by denying the risk but by structuring around it. Their work schedule, during the early days of Oscilar, began at 5 a.m. after a few hours of sleep, carving out the pre-dawn hours for focused work before their child woke up. The discipline is monastic, and it reflects a philosophy that treats the startup not as a lifestyle but as a finite sprint demanding total commitment.
Kulkarni's background — over a decade at Facebook, deep expertise in infrastructure at global scale — is complementary to Narkhede's in ways that mirror the Confluent co-founder dynamic. She is the external face of the company, the strategist, the fundraiser (even when the funds come from her own bank account). He is the infrastructure architect, the builder of systems that must work at scale and under adversarial conditions. The division of labor is legible and, by all evidence, effective.
I saw a lot of use cases on real-time data, and the fastest growing and the most interesting one was fraud detection and risk decisioning. And that is what led to Oscilar.
What the Second Time Teaches
Narkhede has spoken publicly about the difference between founding a company for the first time and the second. The analogy she favors, borrowed from fellow repeat founder Dennis Pilarinos, is of walking through a dark room: "The second time, you roughly know where the furniture is, but you'll still bonk your toe on a few things."
The substantive differences in her approach to Oscilar reveal what she learned — and what she rejected — from Confluent. At Confluent, the company's zero-to-one phase was compressed because the technology (Kafka) already existed and had adoption. At Oscilar, there was no pre-existing open-source project to build on; the product had to be created from scratch, which meant a longer, more uncertain path to product-market fit.
To manage this uncertainty, Narkhede developed what she calls a "proactive research sprint" — a structured process for testing whether a problem is real, whether customers will pay to solve it, and whether her team can build a solution that is demonstrably better than the alternatives. The method is empirical rather than instinctive: she talks to dozens of potential customers, maps the competitive landscape, identifies the specific gaps that existing solutions fail to address, and only then begins building.
The approach to go-to-market is also different. At Confluent, the open-source distribution model created a natural top-of-funnel: developers adopted Kafka, and sales teams converted their companies into enterprise customers. At Oscilar, there is no open-source funnel. The GTM motion is more traditional — outbound sales, industry conferences, partnerships — but informed by the lesson Narkhede learned at Confluent about the importance of founder-led sales in the early stages. She is personally involved in closing deals, a practice she considers non-negotiable for any founder in the first two years.
The other lesson she carries is about monetization architecture. At Confluent, the transition from licensed software to consumption-based cloud pricing was strategically transformative but operationally painful. At Oscilar, the pricing model was built for consumption from the beginning — no retro-fitting required.
The Pattern of the Father's Stories
There is a thread that runs through Narkhede's story that is easy to overlook if you focus only on the technology. It is the thread of representation — the idea that seeing someone who looks like you do the impossible thing is itself a technology for making the impossible possible.
Her father's stories of Indira Gandhi, Indra Nooyi, and Kiran Bedi were not fairy tales. They were case studies. Each woman he chose had accomplished something that was supposed to be impossible for a woman in India: leading the nation, running a Fortune 500 company, commanding a police force. The stories taught Narkhede not merely that women could achieve, but how they had done so — by being, as she would later put it, "a little deaf" to the voices that told them they couldn't.
"Being a little deaf helps quite a bit," Narkhede has said. "It's a survival strategy." The deafness is not ignorance of bias — she is forthright about its existence, noting that "men tend to be assessed by their future potential, and women assessed by past experience" — but a deliberate refusal to let bias dictate her actions. "Develop the mental strength to fight biases and sexism," she has advised other women engineers. "Do not quit."
The advice is both blunt and nuanced. She does not pretend that the obstacles are imaginary, which would be condescending. She acknowledges that "some of it actually does exist in varying proportions in different cultures." But she insists that the response must be strategic, not emotional: write down what you want to say before a negotiation, anticipate objections, make it okay to hear "no" a couple of times, and if the "no" comes without clear and actionable feedback, leave. The negotiation advice has the crisp specificity of someone who has rehearsed these conversations many times — as she has, by her own account — and who recommends Chris Voss's
Never Split the Difference as a tactical manual.
In 2022, she received the Technology Entrepreneurship Abie Award from AnitaB.org. In 2024, Georgia Tech inducted her into both its College of Computing Hall of Fame and its 40 Under 40 alumni class. The recognition was institutional and formal, but its meaning was personal: the girl who had read about Indira Gandhi in Pune had become the kind of figure that fathers in Pune might now tell their daughters about.
The Nervous System Thesis
Narkhede has used the same metaphor across both of her companies, separated by seven years: the nervous system. "We view our technology as a central nervous system for companies that aggregates data and makes sense of it within milliseconds, at scale," she said of Confluent in 2017. At Oscilar, the language is nearly identical — a real-time risk intelligence platform that serves as the nervous system for a financial institution's entire risk operation.
The metaphor is not accidental. It encodes a specific worldview about the nature of modern enterprise technology: that the fundamental problem is not computation or storage but connection — the ability to move information from where it is generated to where it is needed, in real time, at scale, with sufficient context to drive decisions. At Confluent, the connections were between applications and data stores. At Oscilar, they are between risk signals across the customer lifecycle. The underlying abstraction — data in motion — is the same.
This consistency suggests that Narkhede is not a serial entrepreneur in the conventional sense, chasing opportunities across unrelated domains. She is a systems thinker who has identified a single deep problem — the inadequacy of batch-oriented, siloed data architectures in a world that demands real-time, unified intelligence — and is building companies that attack different manifestations of it. Confluent was the infrastructure layer: the plumbing. Oscilar is the application layer: a specific, high-value use case built on the principles that Confluent established.
The bet is that risk management — encompassing fraud detection, credit decisioning, identity verification, compliance monitoring, and anti-money-laundering — will increasingly converge into a single, AI-driven platform. If Narkhede is right, Oscilar will be positioned at the center of a market that touches every financial institution, fintech, and e-commerce platform in the world. If she is wrong, the $20 million she and Kulkarni invested will have bought them an expensive education. Either outcome, she seems to accept with the equanimity of someone who has already proven she can build a company from nothing to a $9 billion public listing.
The Image That Resolves
At the Kafka Summit in London in 2019, Narkhede opened her keynote by describing how the role of technology in business had fundamentally changed. "In the past, enterprises would think of tech as a support function, the CIO was thought of as a technical leader," she said. "But today technology is the business and the CIO is the business leader." She paused. Then she told the audience that 60 percent of Fortune 100 companies were running Apache Kafka. "Even today, the new companies that get created digital-wise use it as a basis for their business."
It was a factual observation delivered without triumphalism, but the subtext was unavoidable. The technology she had helped invent in a conference room at LinkedIn — the open-source project she had shepherded from an internal tool to an industry standard, the distributed commit log that processes trillions of messages for companies whose names are synonymous with the modern internet — had become infrastructure so fundamental that new companies built on top of it the way they built on top of electricity. And the woman who co-created it was standing on a stage in London, in front of an audience of engineers from around the world, telling them what came next.
Somewhere in Pune, a mechanical engineer's daughter had done the impossible thing.
