Autonomous AI agents are ushering in a new era for verification by offloading work traditionally performed by CAD teams and expert DV/FV engineers. These autonomous agents use specifications, design‑related data, and verification context—combined with seamless access to Cadence tools and internal data models—to autonomously generate verification collaterals and execute verification tasks with unprecedented efficiency, consistency, and quality.

This session will introduce the Cadence Agentic Verification, a coordinated multi-agent system orchestrated with Cadence’s verification platforms to achieve full-scale DV and FV goals. These autonomous agents generate comprehensive verification environments—or upgrade existing ones—incorporating all required verification collateral, such as UVM testbenches, SystemVerilog assertions, and verification plans, with full traceability back to the product specification. Central to this approach is a mental model derived from the product specification and RTL, that captures design intent, functional behavior, interface protocols, and boundary conditions, providing a rich contextual foundation that guides agent reasoning. Engineers can refine this model using natural language, while it simultaneously accelerates agent autonomy and deepens the verification team’s understanding of the design.

We will also showcase advancements in Cadence Verisium AI that extend autonomy into regression and debug, including intelligent testcase selection, regression orchestration, AI‑driven failure triage, bug‑pattern prediction, and context-aware waveform root‑cause analysis.

By combining autonomous agents with the Cadence data models, engines, and domain expertise, our solutions dramatically improve engineering efficiency, accelerate verification signoff, and enhance silicon quality—sparing verification teams the time to focus on architecture and innovation