Israeli AI Startup Plurai Emerges from Stealth with $13 Million Seed Round
Israeli AI startup Plurai launched out of stealth with a $13 million seed round led by Team8. The platform automates testing and optimization for enterprise AI agents using simulation-driven methodologies.
Israeli artificial intelligence startup Plurai has emerged from stealth mode, announcing a $13 million seed funding round led by Team8, with participation from Velocity, U&I, Mercer, and prominent industry angel investors. The company's platform enables automated testing and optimization of AI agents prior to their deployment in real-world production environments, ensuring continuous reliability and quality monitoring.
Backing from Industry Leaders
The funding round attracted notable tech executives and experts, including Udi Mokady, founder of CyberArk, and Riah Miara, Senior Director of Google Models at Google Cloud and former head of model evaluation at OpenAI. Additionally, Dr. Eyal Karmi, CTO of Microsoft 365 Copilot, serves on Plurai's advisory board. Founded in 2025 by CEO Dr. Ilan Kedar and CTO Dr. Elad Levy, both computer science and mathematics PhDs with over a decade of experience in large-scale AI deployment, the startup draws on their background in autonomous vehicle evaluation systems, such as Waymo, developed during their tenure at Nexar.
Simulation-Driven Reliability
Transitioning AI agents from a demo phase to a reliable enterprise production environment remains a core industry challenge. While building a demo takes hours, actual deployment exposes agents to unpredictable users, sensitive data, and complex edge cases. Plurai applies autonomous driving simulation principles, creating digital twins of both the agent and its operational environment. Operating within a secure sandbox, the platform tests agents against thousands of scenarios to detect malfunctions, hallucinations, unsafe actions, and tool failures prior to launch.
Beyond identification, Plurai automatically optimizes the agent and re-tests performance to guarantee stability. The technology also allows organizations to replace expensive large language models with smaller, task-specific models, significantly reducing operational costs. Recently, the company announced a collaboration with Nvidia and Nebius, demonstrating a technology that converts open-source AI models into task-specific solutions within 20 minutes at a fraction of the cost of leading models from OpenAI and Anthropic.
