Build a managed experimentation workbench that connects to event streams, support logs, customer conversations, and marketing systems to define experiments, track metadata, monitor bad cases, and summarize results with AI-assisted analysis. The product would support both online A/B tests and offline evaluation sets so teams can pressure-test AI-driven workflows before making heavier engineering investments.

AI agents and AI-enabled GTM/support workflows are moving into production, but companies still need reliable experimentation layers around them. Multiple postings explicitly mention A/B testing, causal inference, offline evaluation, logging, and rapid test iteration as active hiring needs.