Accounting Quality Architect for AI Evaluations

Obsidian; San Francisco (CA); On-site; USD 120,000 - 170,000; Full time

Job summary

Mercor is partnering with a leading AI research organization to engage experienced accountants for a project that evaluates how well AI systems perform real‑world accounting work. You will design task‑specific grading criteria and define what excellent work looks like.

You will design task‑specific grading criteria, score both AI‑generated and human work samples, and write detailed justifications to ensure scores are reproducible and defensible.

Qualifications

• 5+ years of professional accounting experience.
• Background in audit, assurance, or advisory at leading global accounting firms, or in senior accounting roles at large public companies.
• CPA or equivalent certification strongly preferred.
• Deep fluency in US GAAP application, month-end close, account reconciliations, financial statement preparation and disclosures, and audit-ready workpapers.
• Exceptionally strong written communication, with the ability to explain exactly why a piece of work does or does not meet professional standards.
• Detail-oriented, consistent, and comfortable having your judgment reviewed and calibrated against peers.
• Prior experience with AI training, evaluation, or human-data projects is a strong plus.

Responsibilities

• Define task-specific grading criteria for real-world accounting deliverables (reconciliations, close packages, financial statements and disclosures, workpapers, technical accounting memos).
• Score AI-generated and human work samples against those criteria, with detailed written justifications for every score.
• Apply consistent, evidence-based judgment so that scores are reproducible and defensible.
• Incorporate structured feedback from senior reviewers and iterate quickly on your work.

Skills

Advanced accounting

US GAAP

Audit/assurance

Strong writing

Attention to detail

Education

CPA or equivalent

B.S./B.A. in accounting

Tools

Excel

Job description

Mercor is partnering with a leading AI research organization to engage experienced accountants for a project that evaluates how well AI systems perform real‑world accounting work. You will design task‑specific grading criteria and define what excellent work looks like.

You will design task‑specific grading criteria, score both AI‑generated and human work samples, and write detailed justifications to ensure scores are reproducible and defensible.