/zendesk

Zendesk is a global customer experience platform used by more than 100,000 companies. I joined the Berlin AI team to work on agentic AI products that automate customer support at scale.

I designed and built an AI agent security framework that became the pre-deployment security gate for Zendesk's AI agents ahead of their public GA release. It covers more than 1,500 attack scenarios, including prompt injection, jailbreaks, policy violations, and adversarial inputs, and is standardized across agent releases.

I also designed an LLM-as-a-Judge online scorer that reaches 92% agreement with human raters and deployed it in Braintrust to monitor production agent quality continuously. This replaced manual QA sampling with an automated, always-on quality signal.