AI agent security, built on PSE. Vid sees what reaches your agent; Perun stops the attack. Made in Toronto.
Toronto, Ontario
Zerava Technologies Inc. protects AI agents on our customers' own computers. An agent that reads documents, sends email, moves files or pays can be talked into things by one hidden line in what it reads. We stop that. Our products stand between the agent and its model, on your hardware, with no cloud and no telemetry. Vid is free and open source (GPL-3.0). It shows what really reaches your agent: hidden and encoded text, forged roles, secrets leaving in a reply. It undoes 42 kinds of disguise. It shows; it blocks nothing. Perun is the paid guard. It decides on every call whether to let it through, clean it or stop it, covers eight of the nine ways agents get attacked, and sends only the undecided cases to a human, with the reason. Every decision goes into a signed record that stays with you. Ubuntu and Windows. Mapped to the OWASP Top 10 for LLM Applications. Both run on PSE, our open safety engine. The theoretical framework is published in full; the free tools are open source: Pivot Benchmark (where a model breaks under social pressure, and whether it recovers), PSE Analyzer (the mathematics of a local model's weights, in plain words) and PSE Triglav (agents that keep each other honest, in preview). Toronto, Ontario. Founded 2025 by Andrej Ristić (Chief Researcher) and Dhiraj Sah (Chief Software Engineer). Member of NVIDIA Inception. Design partners: three places, 60 days at no cost, half price for the first year. Talk to a founder: info@zeravatech.ca · zeravatech.ca
Remote · Canada
Sell Perun and Svarog AI security software to small and mid-sized businesses and to MSPs and IT-service firms. Explain the products to prospective buyers, conduct sales calls, and develop customer relationships that can generate ongoing renewal commissions.