Noam Awadish
He was Head of Initial Access at Israel's Unit 8200, which is a polite way of saying he spent a decade getting into places he was not invited. Now he runs a company that decides who gets to be believed on a Zoom call.
The LeadThe impersonation economy has a new adversary
Noam Awadish runs imper.ai, a cybersecurity company that in December 2025 announced $28 million in funding and a public launch. Redpoint Ventures and Battery Ventures led the Series A. The pitch is simple to state and hard to build: the deepfake did not create a new attack, it just made the oldest attack cheaper. Somebody calls, sounds convincing, asks for something. Historically the ask was routed through a person's judgment. Awadish's argument is that in 2026 the judgment is losing.
The interesting move is what imper.ai does not do. It does not sit on the video stream and try to spot the pixel that betrays a synthetic face. "We don't want to get into an AI arms race," Awadish told Fortune when the company launched. Instead the platform reads what an attacker cannot easily fake even with an unlimited GPU budget: the fingerprint of the device on the other end, the way its network is behaving, and how the session compares to the private context of the organization that owns it. If you accept the premise that deepfakes are already good enough, this is the only place left to look.
imper.ai is agentless and covers the surfaces where enterprises actually talk to each other and to strangers: Zoom, Teams, Slack, WhatsApp, applicant tracking systems, the IT help desk. The through-line is that every one of these tools was built for convenience first and identity verification second. Awadish is selling the retrofit.
He founded the company with Anatoly Blighovsky, imper.ai's chief product officer, who was head of two cyber divisions and CISO at Unit 8200, and Rom Dudkiewicz, chief technology officer, a vulnerability researcher and section leader at the same unit. Three co-founders, one alma mater. There is a long tradition of Israeli security startups arriving with this kind of pedigree; imper.ai is the latest.
I think that people don't understand that most of the major breaches start with social engineering.— Noam Awadish, to Fortune, December 2025
Section II · OriginFrom Unit 8200 to Mobileye to imper.ai
Awadish's public biography compresses into a single sentence and expands into a career. He spent more than a decade in Unit 8200, the Israeli military's signals intelligence organization, rising to Head of Initial Access. Then he moved to Mobileye - the Jerusalem company that built the front-facing camera systems inside a majority of the world's assisted-driving fleets - where he was Chief of Staff to the executive vice president running the Autonomous Vehicles group. Somewhere in the middle he earned a Bachelor of Science from The Hebrew University of Jerusalem.
Chief of Staff is one of those titles that means different things in different companies. Inside a public-company operating unit at the scale of Mobileye's AV program, it means calendar, headcount, quarterly targets, and the internal politics of getting anything shipped in an organization whose customers are car OEMs. It is the sort of job that trains a person to run a startup - which is presumably part of why he is now running one.
The pivot from autonomous vehicles to autonomous impostors is less strange than it looks. Both are problems of trust at a distance: the AV stack has to decide, in fractions of a second, whether the shape on the road is what it claims to be. imper.ai has to decide, on roughly the same time scale, whether the face on the call is who it claims to be. The engineering is different. The question is the same.
Illustrative. Exact tenures are approximate.
Co-Founders
Anatoly Blighovsky — CPO. Former head of two cyber divisions and CISO at Unit 8200.
Rom Dudkiewicz — CTO. Vulnerability researcher and section leader at Unit 8200.
Section III · The BetWhy he refuses to play the detection game
The obvious way to build a deepfake company is to build a better deepfake detector. Look at the pixels, listen to the phonemes, compare against a training set of known synthetic outputs, ship a confidence score. Awadish thinks this loses. Generative models improve faster than detectors, and the false-positive rate on any content-based system is a tax that no enterprise buyer will pay for long.
So imper.ai is designed around signals that a synthetic voice on a call does not control. What device is the person using. Is the microphone the microphone it claims to be. Does the network path match a claimed identity. Does the timing of the session line up with what the organization's calendar system already knows to be true. This is what the company means when it talks about "digital breadcrumbs attackers cannot fake." The market pitch is quieter than the deepfake headlines but the math is cleaner.
Device telemetry
A synthetic speaker still runs on some machine, in some browser, on some OS. That stack has a fingerprint.
Network diagnostics
Route, latency, geography, provider. The plumbing lies less often than the pixels do.
Behavioral markers
Session shape, timing, cross-referenced against what the customer already knows about its own people.
We want to build a platform that safeguards the entire communication space. It's not something small.— Awadish on imper.ai's scope
Section IV · LedgerCareer timeline
Section V · In His Own WordsFour quotes, in the order he said them
On breaches
"Most of the major breaches start with social engineering."
On strategy
"We don't want to get into an AI arms race."
On scope
"It's not something small. It's not a plugin one of the giants is going to build."
On the market
"AI-driven impersonation has become one of the biggest drivers of financial loss and reputational risk for enterprises."
Section VI · DetailsSmall facts, arranged sideways
The founding team's edge, in Awadish's telling, comes from having spent a decade on the wrong side of the firewall. Head of Initial Access is not a defensive title. It is the person tasked with getting inside. When you spend that long thinking like an intruder, you develop opinions about which defensive products are theater and which are real. imper.ai is the second kind, or so its founders would like you to believe.
The company's name is a compression of the word "impersonation" - which is what its software is designed to prevent - and the two letters, ai, which the market currently demands. There is not much room for interpretation. It does what it is called.
imper.ai already had 37 employees by the time it announced its funding. That is a lot of people for a company that had been operating quietly for less than a year. The plan announced with the launch was to double the R&D group and triple the U.S. go-to-market team, which is the polite venture-backed phrase for "we are hiring salespeople in New York this quarter."
Section VII · FAQQuestions people actually ask
Who is Noam Awadish?
Co-founder and CEO of imper.ai, a New York-headquartered cybersecurity company that detects AI-driven impersonation and social engineering attempts in real time.
What did he do before imper.ai?
Chief of Staff for Autonomous Vehicle Programs at Mobileye, following more than ten years in Israel's Unit 8200 where he was Head of Initial Access.
How much has imper.ai raised?
Approximately $28 million in total - a roughly $6.5M seed round followed by a Series A of about $21.5M announced in December 2025.
Who are the investors?
Redpoint Ventures and Battery Ventures led the Series A, with participation from Maple VC, Vessy VC, and Cerca Partners.
What does imper.ai actually do?
Its agentless platform reads device telemetry, network diagnostics and behavioral markers across Zoom, Teams, Slack, WhatsApp, ATS and help desk tools to flag impersonation without scanning content for artifacts.