Kevin
Tian.

Somewhere on the internet right now, a fake version of a company you trust is running an ad, cloning a login page, or spinning up an account that looks exactly like the real thing. Kevin Tian's job is to find it and delete it before you ever click. He is the co-founder and CEO of Doppel, an AI-native platform built for one unglamorous, endlessly renewable problem: people get fooled, and now machines can fool them at industrial scale.

The strange part is where Tian came from. Before cybersecurity, he wrote software for flying cars. At Uber and Lyft he built the real-time dispatch systems that decide, in milliseconds, which driver goes where - and he worked on the aerial and autonomous vehicle programs that were supposed to lift ride-hailing off the road entirely. It was distributed systems at brutal scale. It had nothing to do with phishing. It turned out to be exactly the right training.

"The most dangerous thing about AI is how easy it is to impersonate anyone."

Tian is a Georgetown graduate who started his career as a junior engineer at FiscalNote and interned in data science at Facebook, both in 2015. Then came the ride-hailing years. At Lyft he met Rahul Madduluri, who had worked on UberPOOL - the carpooling logic that had to match strangers into the same moving vehicle without either of them noticing the seams. The two engineers liked hard matching problems. They would soon find a much harder one.

The bet nobody wanted in 2022

Madduluri happened to be roommates with a head of research at OpenAI. That gave the pair an early, unfiltered read on where large language models were heading - not the hype version, the engineering version. And what they saw worried them. Generative AI was about to make deception cheap. A convincing fake voice, a cloned website, a personalized phishing message, a deepfake on a live video call: all of it was collapsing from expensive-and-rare to free-and-infinite.

So in 2022 they started Doppel. The thesis was blunt: social engineering was becoming the number one source of breaches and the number one source of cyber-insurance claims, and it was about to get much worse. Most security tools guarded the network. Tian wanted to guard the thing attackers were actually exploiting - trust.

"Generative AI is normalizing deception - making deepfakes, fake ads, and cloned voices cheap."

Start small, then eat the map

Doppel began where impersonation was most feral: crypto. Scam tokens, fake support accounts, and lookalike sites were bleeding crypto companies daily, so that is where the product cut its teeth. Once the engine could keep up with that volume and velocity, Tian pushed it up-market into enterprises and Fortune 500 brands. The company now protects names like Coinbase, Ramp, Notion, and the law firm Orrick.

The core idea is a "threat graph." Doppel scans the open internet - fake accounts, malicious ads, spoofed domains, fraudulent messaging - and then connects the dots between them: this phone number links to that IP, which links to this advertiser account, which links to a dozen more fakes. Kill one node and you often collapse a whole network. When Notion got hit by malicious ads and cloned accounts, Doppel's agents shut down thousands of threats and knocked down the follow-on attacks that usually come next.

From co-pilot to autonomous agents

The part Tian is proudest of is also the part that makes traditional security operations centers nervous. Doppel moved from a "co-pilot" model - AI that helps a human analyst - to autonomous AI agents that review millions of alerts a day and act on them. In practice, that means the software now does the work of a tier-one analyst, at a volume no human team could match, and, Tian argues, with better accuracy.

To stay ahead, his team also builds the attacks. Doppel's simulation engine works, in Tian's words, "very much like vibe coding": you give it a prompt, and an AI agent constructs a phishing campaign that targets a specific person on their phone - maybe a call, then an SMS right after to confirm the details. The newest and most unsettling capability, he says, is real-time, synchronous deepfake conversation. Doppel builds the nightmare on purpose so it can recognize it in the wild.

"The limitation of your organization will be your imagination and ambition."

A very fast eighteen months

The market agreed with the thesis. In May 2025 Doppel raised a $35M Series B at a $205M valuation, led by Bessemer Venture Partners. Just six months later, in November 2025, it closed a $70M Series C at a valuation north of $600M - again led by Bessemer - bringing total funding to $124M. The Series C roster reads like a security-industry endorsement and a pop-culture surprise at once: CrowdStrike founder and CEO George Kurtz wrote a personal check, NTT DOCOMO Ventures and Aurum Partners (the fund tied to the San Francisco 49ers' owners) joined, and a bench of WNBA champions - Nneka Ogwumike, Breanna Stewart, and Kelsey Plum - came in through the a16z Cultural Leadership Fund.

Under the headlines were the numbers Tian cares about: roughly 3x year-over-year ARR growth, a 5x jump in Fortune 500 customers, a customer base past 200, and a team that more than doubled. "This signals confidence in the path we're on," he said of the raise, "and reinforces our commitment to driving rapid growth."

What he is actually building

Ask Tian where this goes and the answer is bigger than takedowns. He wants Doppel to be the default platform for social engineering defense - the layer companies reach for by reflex - and, beyond that, to help people verify whether what they are seeing online is real at all. In a world where a voice, a face, and a brand can all be forged for pennies, that is less a product category than a public utility. His leadership creed is compact: transparency, vision, energy. His constraint is even simpler. Imagination and ambition. Everything else, he seems to believe, is just distributed systems.