- Own end to end product outcomes for AP AI Specialist end to end
- Build, don't just spec: prototype agent behavior, stand up demos, validate feasibility with engineering before it's a roadmap item
- Take an ambiguous problem (a customer pain point, an eval failure, a platform constraint) and figure out the right shape of the solution without waiting for it to be handed to you fully formed
- Define what "valuable" looks like for agent behavior — success criteria, eval scenarios, confidence thresholds — and use eval methodology as a real lever for product decisions, not a checkbox
- Partner with the platform/orchestration team on architecture tradeoffs (agent design, latency, oversight vs. automation), with enough technical depth to push back when something doesn't hold up
- Translate technical tradeoffs into plain language for engineering, architects, and execs — get in front of real users early to validate before building further
- Thrive in ambiguity: create structure where none exists, and pull the work forward through clarity and credibility rather than process