Company
Generative AI reached critical mass inside high-stakes assessments in the last twenty-four months. Legacy proctoring — built on webcam behavior analytics and browser lockdowns — has no answer. So we built the first network-layer proctoring platform: a 30-second user-space enclave that prevents AI exfiltration instead of trying to detect it after the fact.
Origin
In 2024, Sanjay was studying for the GMAT when his feed started serving proxy-ring ads — paid services that would sit the exam for you. He pulled the thread with Akshay, who had spent a decade on the offensive side of security and had already noticed the same tooling leaking into hiring pipelines: invisible overlays, local LLM helpers, remote shells disguised as productivity apps.
The assessment industry had optimized for the wrong threat model. Webcams and browser lockdowns were designed for a world where the cheating tool was a friend in the next room or a Google tab. The new tool was a 13-billion-parameter model running locally on-device, invisible to the camera, invisible to the browser, invisible to the operating system's install registry.
Divya — an AI safety and red-team engineer — had spent years reverse-engineering the overlay architectures now being productized as cheating tools. Sudhakar had nine years inside enterprise security organizations, watching how networking primitives could be repurposed for defense. We kept arriving at the same conclusion: the only place you can reliably interrupt the AI-assistance loop is the network layer. Below the browser. Above the OS. Ephemeral. Deployable in under a minute on any candidate laptop, with no driver install and no admin permission.
We incorporated as a Delaware C Corp and shipped the first version. Twenty-five attack vectors neutralized in the MVP. A pilot at an Ivy League research university. One hundred-plus interviews with assessment and talent leaders to validate the buyer problem. The company exists because the next generation of exams can't be secured with the previous generation's tools — and nobody else was building what the math required.
Team
Five people. Each chosen because the network-layer thesis only works if the room combines offensive security, AI red-teaming, enterprise networking, and assessment-market GTM in one hallway. Backgrounds below.
CEO & Founder
Ten-plus years in offensive cybersecurity. Earned $300,000+ in bug bounties across Fortune-500 programs. Co-founded Upskillfinder.com. Holds 30+ industry certifications spanning red-team operations, network security, and cloud infrastructure. Cornell-trained. Leads Aiseptor's technical architecture and sets the threat model.
Chief AI Officer
AI safety and red-team engineer. Reverse-engineered the overlay architectures now productized as the leading AI-cheating tools — before those tools reached public distribution. Runs Aiseptor's adversarial-AI research, including detection of local LLM inference, model fingerprinting, and GPU-memory anomaly analysis.
Chief Product Officer
Nine years inside enterprise cybersecurity organizations at the intersection of AI, networking, and zero-trust architecture. Owns Aiseptor's product surface — from the enterprise dashboard and proctor console down to the agent lifecycle on Windows and macOS.
Chief Business Officer
Five-plus years across AI, cloud, and enterprise go-to-market. Spotted the proxy-ring ad economy first-hand while preparing for the GMAT — the origin story of the company. Runs Aiseptor's partnerships, platform embedding deals, and enterprise relationships.
Public LinkedIn URLs available on request — contact hello@aiseptor.com.
Advisors
A small bench, chosen for depth in the two domains the thesis depends on: the global assessment industry and offensive cybersecurity.
Advisor
Chairman of the eAA Assessment Association (UK). Brings three decades of global assessment-industry perspective — standards bodies, examining boards, and the regulatory gravity around high-stakes credentialing.
Advisor & Board Member
Ranked #1 Ethical Hacker of India on Google. Founder and CEO of Secure Your Hacks. Senior Consultant, Cybersecurity to India's Ministry of Home Affairs. Anchors the offensive-security discipline that sets Aiseptor's detection ceiling.
Technical Advisor
Founder and CEO of Infinite Potential. Delivered $20M+ in technical projects for the Indian Army, Navy, and Air Force. Brings large-scale deployment, sovereign-grade security, and operational resilience experience to Aiseptor's roadmap.
Where we are
Honest stage disclosure. We're alpha — users on production, no revenue yet, bootstrapped roughly $20K to date. The signal that matters is what we've built, who's validated it, and who's signed an LOI.
C-suite assessment and talent-acquisition executives interviewed to validate the buyer problem before a single line of enterprise feature code shipped.
Partnership LOI with a leading assessment platform at $30M ARR and 5,000+ enterprise customers. Integration scope locked.
Two pilot runs with 100+ candidates at an Ivy League research university — real assessments, not simulations.
Distinct AI-cheating and exfiltration attack vectors identified, reverse-engineered, and neutralized in the MVP.
Ephemeral user-space network enclave — deployable in under 60 seconds, no kernel drivers, no persistent install.
Shipped to production on Windows and macOS. Transparent per-session pricing (see /pricing). Pre-revenue by design.
Why now
Three forces converging in a narrow window. The category is being recreated; we intend to author the primitive it runs on.
Invisible overlay assistants and on-device LLMs hit mainstream distribution between 2023 and 2025. Every proctoring product architected before that window is solving the wrong problem — and retrofitting won't close the gap, because the cheating tool now lives below the browser.
Building a user-space network enclave that deploys in 30 seconds, with no driver install, requires deep offensive-security knowledge — WireGuard internals, DNS rewriting, nftables policy, process introspection, GPU-memory telemetry. That skill stack is absent from the teams that built the current assessment-security category.
Palo Alto, CrowdStrike, and Zscaler architect for managed fleets and 30-day deployments. A candidate's personal laptop for a 90-minute exam breaks every assumption in their model. The innovator's dilemma leaves the seam open.
Press
For press inquiries, quote requests, or brand-asset downloads, visit our press kit. Journalists covering the AI-cheating beat are welcome to request a technical briefing with the founding team.
Coverage
We're early — no press mentions to list yet. When coverage lands, it will appear here and in the press kit. In the meantime, the research report is the best starting point for context on the market and threat model.
We read everything that lands in the inbox. Partnership conversations route to the founding team directly. Hiring is selective — we only open roles when the work is real and the ownership is meaningful.