Senior Machine Learning Engineer

Zendesk
Melbourne, New South Wales
Workplace: HybridFull timeFunction: Data Science & Machine LearningSkills: []

Build and harden the Resolution Platform’s agent runtime that lets AI agents perceive, reason, and act to autonomously resolve Zendesk customer service tickets. Work on large-language-model planning, multi-agent coordination, memory structures, evaluation systems with regression detection, and enterprise guardrails against misuse, prompt injection, and hallucination loops. Own both the research and systems needed to develop domain-specialized models via RL and production feedback loops.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
Zendesk
Zendesk
1 month ago

Senior Machine Learning Engineer

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 25 days agoStatus: Live

Job Summary

Build and harden the Resolution Platform’s agent runtime that lets AI agents perceive, reason, and act to autonomously resolve Zendesk customer service tickets. Work on large-language-model planning, multi-agent coordination, memory structures, evaluation systems with regression detection, and enterprise guardrails against misuse, prompt injection, and hallucination loops. Own both the research and systems needed to develop domain-specialized models via RL and production feedback loops.
Location: Melbourne, New South Wales
Workplace: Hybrid
Employment Type: Full time
Job Function: Data Science & Machine Learning
Seniority: Mid level

Key Responsibilities

  • •Design and advance agent architectures for planning, tool dispatch, memory, and error recovery used to resolve customer service tickets without a human in the loop.
  • •Improve multi-agent coordination, including agent delegation patterns using the Agent-to-Agent (A2A) protocol.
  • •Develop and own reinforcement learning infrastructure for domain-specialized customer service resolution models, including reward curricula and feedback loops.
  • •Harden automated evaluation systems with multi-turn analysis and trajectory insights, integrating quality gates into CI.
  • •Implement enterprise guardrails (supervisor patterns, capabilities-based access control, and output validation) to mitigate tool misuse, prompt injection, and hallucination loops at scale.

Key Requirements

  • •5+ years building production ML/AI systems with hands-on experience in agent architectures (planning, tool dispatch, memory, error recovery).
  • •Strong evaluation instincts: build internal evals to close gaps between public benchmarks and production performance.
  • •Optional: depth in RL for language models, including reward shaping and online/offline tradeoffs.
  • •Python and PyTorch fluency, with familiarity with at least one agent framework and good judgment on when to build custom.
Tech Stack:PythonPyTorchLangChainLarge language modelsMulti-agent coordinationRLReward shapingBraintrustA2A (Agent-to-Agent protocol)CI

Company Brief

Zendesk
Provides cloud-based customer service and engagement software, including help desk, ticketing, chat, and CRM tools that enable businesses to manage customer support, self-service, and omnichannel communication at scale.
Industry: Enterprise Software
Company Size: Enterprise (1,001+ employees)
Revenue: USD 1B+
Growth: Established Company
Valuation: Decacorn (USD 10B+)
Funding: Private Equity Backed
Headquarters: San Francisco, United States
Founded: 2007
Glassdoor
Glassdoor: 4.0
WebsiteLinkedInGlassdoor