Women Building the AI Era, Part 4- Neha Narkhede

In 2010, LinkedIn had a problem most of its users never noticed. Every click, every profile update, every new connection generated a small burst of data, and the company's systems couldn't move all of it around fast enough to make features like real-time recommendations actually work.

Neha Narkhede, then an engineer on LinkedIn's data infrastructure team, helped build the fix: an open-source tool called Apache Kafka, designed to move enormous streams of data in real time without falling behind. LinkedIn gave it away for free.

Fifteen years later, in December 2025, IBM paid 11 billion dollars for the company Narkhede started to commercialize that same idea, not because Kafka had changed, but because AI agents had arrived and suddenly every company building one needed exactly what Kafka does.

A Problem Nobody Had Solved Yet

By the early 2010s, LinkedIn wasn't alone in needing to move data faster than traditional databases were built to handle. Narkhede and two colleagues, Jay Kreps and Jun Rao, built Kafka specifically for that gap: a system that could take a constant flood of events and route them wherever they needed to go, in something close to real time.

Think of the difference between a mailbox and a live phone line. A traditional database is a mailbox: you check it when you get around to it, and by the time you look, some of what's in there is already hours old. Kafka behaves more like a phone line that never hangs up, moving each new piece of information the moment it happens, so the systems on the other end are always working with what just occurred, not what occurred that morning.

They open-sourced it early, a decision Narkhede has said was deliberate. Foundational plumbing like this only becomes an industry standard if other engineers can adopt it freely, and Kafka did. Within a few years, companies far outside LinkedIn were quietly running parts of their business on it: Goldman Sachs used it to move trading information to traders in real time, Netflix used it to feed its recommendation engine, and Uber used it to power the pricing system that adjusts fares as demand shifts.

Did You Know?

Confluent wasn't Narkhede's first attempt to explain Kafka's value to the world; giving it away for free was. She and her co-founders open-sourced Kafka years before starting a company around it, betting that a widely adopted free tool would create more long-term value than trying to sell a proprietary one from day one.

From Open-Source Project to a Company Worth Billions

In 2014, Narkhede, Kreps, and Rao left LinkedIn to found Confluent, built specifically to support and commercialize Kafka for companies that needed more than the free version could offer. The company went public in 2021, and reports at the time put its valuation at roughly 9 billion dollars.

Narkhede stepped down as Confluent's chief technology officer in 2020, staying on the company's board while moving toward new ventures. By the time IBM came calling in December 2025, Confluent had grown into a company that, according to figures released around the deal, was generating more than a billion dollars a year in revenue, with more than six thousand enterprise customers, including a large share of the Fortune 500, running on it.

The Deal That Made the Plumbing the Point

IBM's 11 billion dollar all-cash purchase of Confluent wasn't framed as a data-tools acquisition. It was framed as an AI acquisition. The company's own announcement described Kafka's real-time data movement as the foundation an enterprise needs before it can safely run AI agents at all, since an agent making decisions on stale or incomplete data is a liability, not a feature.

Picture a fraud-detection agent deciding whether to approve a bank transfer. Working off an account balance from an hour ago, it can wave through a transaction that should have been blocked. Real-time data is what tells the agent what the account looks like right now, not what it looked like before lunch.

That's the quieter argument running underneath the broader story of AI agents learning to act on a company's behalf: an agent is only as useful as the data arriving in front of it in the moment it has to act. Kafka's job, moving that data continuously rather than in slow, scheduled batches, went from a nice-to-have to something IBM was willing to pay a premium for. Confluent's stock jumped nearly 30 percent the day the deal was announced.

Confluent had already been building toward this before IBM came calling. Months earlier, the company launched a product called Confluent Intelligence, designed to stream real-time context straight into AI agents without engineers having to wire up Kafka themselves, and set Anthropic's Claude as the default model running on top of it. The plumbing wasn't just carrying data anymore; it was becoming the delivery system AI agents plug straight into.

Interesting Fact: Narkhede wasn't running Confluent day to day when IBM bought it. She'd already moved on in 2023 to co-found a new company, Oscilar, which uses AI and the same kind of real-time data thinking to catch fraudulent financial transactions as they happen, rather than after the fact.

Building the Layer Underneath, Then Moving On

There's a pattern in Narkhede's career that's easy to miss if you only look at any single company. She doesn't stay to run the thing she built once it becomes an established, large-scale business. She builds the underlying layer, watches it get adopted at scale, and then moves toward the next unsolved problem, in this case AI-era fraud detection at Oscilar, rather than the higher-profile work of training the models sitting on top of that infrastructure.

That's a different kind of AI-era wealth story than the one usually told. It isn't a founder riding one company's valuation upward for a decade. It's someone who built a piece of infrastructure early enough that, by the time the rest of the industry realized how badly it needed exactly that plumbing, she'd already collected the reward and moved to the next problem.

Knowlegic Perspective

It's tempting to tell this as a straightforward founder success story, but the more interesting detail is the timing. Narkhede and her co-founders built Kafka to solve a problem that had nothing to do with AI. Real-time data movement mattered to social networks, banks, and ride-hailing apps years before anyone was building an AI agent that needed to act on fresh information in the moment.

That's exactly why IBM's price tag makes sense. The AI boom didn't create the need for this kind of infrastructure, it revealed how much of it had already been quietly built and adopted, waiting for a reason this valuable to come along. The companies profiting most from AI right now aren't always the ones building the flashiest models. Sometimes they're the ones who built the boring, essential layer underneath a decade early and simply waited.

That's a lesson for how people build careers too, not just how companies build products. The most visible job in tech right now is training or promoting a flashy model. Narkhede's career suggests a quieter bet can pay off just as well: solve an unglamorous problem thoroughly enough that everyone ends up depending on it, then move on before anyone else even recognizes what it's worth.

Neha Narkhede built the plumbing that moves data in real time long before AI agents needed exactly that. She had already moved on to the next problem by the time the rest of the industry caught up to why it mattered this much.

Sources & References

• IBM to Acquire Confluent to Create Smart Data Platform for Enterprise Generative AI, IBM Newsroom (2025)
• Confluent stock soars 29% as IBM announces $11 billion acquisition deal, CNBC (2025)
• Exclusive: Confluent Cofounder Neha Narkhede's New Fraud Detecting Firm Oscilar Emerges From Stealth, Forbes (2023)
• Confluent Intelligence Expands Real-Time Business Data to Enterprise AI, Business Wire (2026)
• Technista Talk: Neha Narkhede - LinkedIn Employee To Confluent Boss, Forbes (2017)