Influencer Program Architect for AI Training Environments

Obsidian; New York (NY); Remote; USD 90,000 - 130,000; Full time

Job summary

Obsidian is building high-fidelity simulated work environments used to evaluate and improve AI agents on real marketing tasks. You will design creator and influencer tasks, specify the documents, dashboards, personas and tool states a realistic task requires, and write the reference answer and the criteria that separate a strong response from a plausible‑but‑wrong one.

Tasks span diagnosis, prioritization, planning, QA and reconciliation, triage and escalation, and multi‑tool orchestration.

Qualifications

• Hands-on practitioner experience running creator or influencer programs.
• Comfortable with creator platforms and paid amplification practices.
• Prior experience building AI training environments, RL environments or simulated case studies is strongly preferred.
• Clear written reasoning — ability to explain why a decision is right.

Responsibilities

• Design task briefs drawn from work you have actually done running creator programs, including what the task is really testing — the planted issue and the decision a strong practitioner should reach
• Specify the documents, dashboards, personas and tool states a realistic version of that task requires
• Write or review the reference answer and the criteria that separate a strong response from a plausible-but-wrong one
• Review AI agent attempts and judge them against your own standard

Skills

Hands-on creator program experience

Creator platform familiarity

Paid amplification knowledge

Clear written reasoning

Experience designing AI training/envir

Job description

Obsidian is building high-fidelity simulated work environments used to evaluate and improve AI agents on real marketing tasks. You will design creator and influencer tasks, specify the documents, dashboards, personas and tool states a realistic task requires, and write the reference answer and the criteria that separate a strong response from a plausible‑but‑wrong one.

Tasks span diagnosis, prioritization, planning, QA and reconciliation, triage and escalation, and multi‑tool orchestration.