Roark

A platform that simulates voice AI agents against hundreds of scenarios before launch and scores every production call on 64+ audio-native metrics, catching issues like mispronunciations and compliance failures. It sells to engineering, data, and operations teams in mid-market and enterprise companies, particularly in healthcare, finance, and customer support. It differentiates by using purpose-built audio models rather than LLM text grading, and integrates with stacks like Vapi and Retell.

Key features

• Simulation testing with hundreds of scenarios
• Post-call analysis with 64+ audio metrics
• Self-improving agents loop
• Load and health tests
• Regression testing
• Red teaming with adversarial callers
• 45 languages and accents
• CI integration for quality gates
• Compliance scoring for healthcare
• Identity verification checks