Zendesk is hiring a Staff ML Engineer for the Agentic AI focus within the AI Agents team, with options to work in Melbourne, Sydney, or remotely in AU. The role involves supporting production AI agents that autonomously resolve customer service tickets across 100k+ Zendesk accounts, planning multi-step resolutions and executing actions via live APIs to close tickets without human intervention. The agent core uses an iterative planning architecture with goal decomposition, a reusable skills registry, execution and evaluation loops, and a self-learning mechanism that turns successful patterns into new skills, with performance matching GAIA-class benchmarks and an internal suite of 158+ scenario tests plus continuous regression checks. They seek to push architecture further, addressing issues like ambiguous goals, memory interference across concurrent sessions, selective skill acquisition, and agent-to-agent delegation, while building domain-specific RL models and a hardened evaluation/CI process plus enterprise-grade guardrails. Requirements include 5+ years building production ML/AI systems with hands-on agent experience, strong evaluation instincts (including internal evals), optional RL expertise for language models, and Python/PyTorch fluency, withZendesk highlighting hybrid work, diversity and equal opportunity commitments and accommodations.
