Plurai raises $13M seed to simulate customers' digital systems and post-train AI agents for higher reliability
Plurai, a startup founded by Ilan Kadar and Elad Levi (both formerly of Nexar), announced a $13 million seed round led by Team8 to build a simulation and post-training platform meant to improve the reliability of AI agents (techcrunch.com). The company creates a digital twin of a customer's knowledge base and tools, maps possible agent flows, generates synthetic edge-case data, and uses those simulations to evaluate and tune agents and to post-train smaller open-weight models tailored to customers. Plurai says tests with Nvidia and Nebius showed the company's models outperforming mid-tier frontier models at substantially lower cost. The company positions this approach as analogous to simulation-driven development used in autonomous driving and claims integration with agentic tools such as Claude Code, Codex, Copilot and Cursor.
Detalhes técnicos
Plurai builds a digital twin of a customer's electronic systems, including knowledge bases and tooling, then enumerates possible agent flows in a graph and generates synthetic data to surface edge cases and failure modes. The platform can be used to evaluate existing agent stacks and to post-train smaller open-weight models that Plurai says outperformed mid-tier frontier models in their tests with Nvidia and Nebius. Plurai markets this as a cost-reduction strategy for organizations that find general frontier models are out-of-distribution for their specific business data (techcrunch.com).
Linha do tempo
-
Plurai raises $13M seed to simulate customers' digital systems and post-train AI agents for higher reliability
ler a análise desta fonte ocultar
Plurai, a startup founded by Ilan Kadar and Elad Levi (both formerly of Nexar), announced a $13 million seed round led by Team8 to build a simulation and post-training platform meant to improve the reliability of AI agents (techcrunch.com). The company creates a digital twin of a customer's knowledge base and tools, maps possible agent flows, generates synthetic edge-case data, and uses those simulations to evaluate and tune agents and to post-train smaller open-weight models tailored to customers. Plurai says tests with Nvidia and Nebius showed the company's models outperforming mid-tier frontier models at substantially lower cost. The company positions this approach as analogous to simulation-driven development used in autonomous driving and claims integration with agentic tools such as Claude Code, Codex, Copilot and Cursor.
Technical Details
Plurai builds a digital twin of a customer's electronic systems, including knowledge bases and tooling, then enumerates possible agent flows in a graph and generates synthetic data to surface edge cases and failure modes. The platform can be used to evaluate existing agent stacks and to post-train smaller open-weight models that Plurai says outperformed mid-tier frontier models in their tests with Nvidia and Nebius. Plurai markets this as a cost-reduction strategy for organizations that find general frontier models are out-of-distribution for their specific business data (techcrunch.com).