Ply provides an engineering-assurance platform for physical AI systems, automatically linking safety and cybersecurity requirements to live engineering evidence as robots evolve. The platform continuously monitors design changes, isolates what needs re-verification, and preserves conclusions whose basis remains valid—so teams maintain a trustworthy safety case without restarting it for every update.
Funding
Funding not disclosed
Founders
Product
Problem
Physical AI systems—robots and autonomous machines—evolve continuously through hardware swaps, software updates, and configuration changes, yet their safety and cybersecurity assurance is typically reconstructed manually from documents, audits, and snapshots. This manual process is slow, error-prone, and quickly becomes stale, leaving safety cases out of sync with the actual engineering state.
Solution
Ply builds a living Assurance Graph that connects engineering artifacts—requirements, controls, tests, and evidence—to the hardware, software, and operating environment they describe. The platform extracts product and environment facts from engineering documents, determines which facts control downstream assurance conclusions, and asks engineers only when something important cannot be established. As changes occur, Ply propagates each decision through the graph, isolating the requirements and tests that need attention while preserving those whose controlling assumptions still hold. Engineers can ask Ply questions grounded in the current engineering state)Skip—the platform reasons over the graph and evidence rather than returning generic chatbot answers.
Target Audience
Ply serves engineering teams and safety managers at robotics companies, autonomous-vehicle developers, and industrial automation firms that need continuous assurance for physical AI products.
Features
- Assurance Graph linking system models, safety/cybersecurity requirements, tests, evidence, and release criteria in one connected structure
- Autonomous extraction of product and environment facts from engineering artifacts, with targeted questions only when critical information is missing
- Change-impact propagation that flags reviews, re-runs, and holds while preserving unaffected evidence and conclusions
- Query interface that reasons over live engineering state and the Assurance Graph instead of providing generic responses
- Standards-to-work conversion turning applicable safety and cybersecurity standards into connected engineering tasks