Agent Test Engineer

The work

Design trajectory tests and tool-call regression packs for multi-step agents. Catch schema drift, wrong tool selection, and argument errors before production traffic — without claiming perfect safety.

You will

• Model agent workflows as testable trajectories with explicit expected tool sequences
• Build regression packs for MCP and function-calling surfaces
• Triage flakes vs real defects with reproducible fixtures
• Document residual risk for delivery leads when agents operate in partial-autonomy modes

You likely have

• 5+ years in software testing or ML engineering with agentic systems exposure
• Experience testing non-deterministic systems with statistical or rubric-based gates
• Strong Python and CI integration skills
• Pragmatic stance on autonomy — test what ships, not slide-deck agents