$60–100/hr · Mercor · Full time, 40 hours a week
Senior accounting professional who trains AI systems to reason rigorously about real audit and accounting work.
What you would do
- Evaluate accounting knowledge work for correctness, logical rigor, and practical applicability.
- Write precise instruction specs and golden solutions that embody what correct accounting judgment looks like.
- Design challenging accounting tasks and evaluation sets based on how practitioners actually work.
- Work with research teams to translate subjective accounting judgment into explicit, teachable rubrics.
- Identify cases where answers sound right but would not survive real-world professional review.
Who they want
- Active professional credential (CPA, CA, CIA, CFE, CMA) with accounting or related degree.
- 4+ years substantive experience at recognized firms, corporate finance, internal audit, or regulatory bodies.
- Specialization in external audit, technical accounting, internal controls, tax, or forensic accounting.
- Senior-level experience (Manager, Senior Manager, Director, or Partner) with real engagement ownership.
- Hands-on use of large language models with judgment to assess reasoning quality.
Main skills
What the interview asks about
1.Spotting accounting mistakes that read plausible
Models often produce textbook-correct answers that ignore real-world constraints; auditors distinguish this from true rigor.
For example: “A model recommends a revenue recognition treatment that follows GAAP mechanically but ignores the specific contract's enforceability and performance obligations; how do you explain why this is wrong?”
2.Technical accounting under different frameworks
Accounting principles vary significantly across GAAP, IFRS, and local standards; experts must apply domain-specific judgment.
For example: “Under IFRS 16, a lessee capitalized a lease at 3.2 million but your firm's analysis shows 2.8 million due to lease modification scope; walk me through the difference.”
3.Designing controls for SOX compliance
Internal audit specialists must understand what constitutes sufficient control design for major transaction cycles.
For example: “A model proposes quarterly reconciliation of intercompany accounts; why is that insufficient for monthly consolidation with 15 subsidiary entities?”
4.Detecting data quality issues in audit work
Spotting which audit observations stem from real risk versus system artifacts or process gaps is critical.
For example: “You're grading a model's audit procedures for a 500M asset base with 2000 individual transactions; it recommends testing only 45 items. What questions does this raise?”
5.Writing clear evaluation rubrics
Rubrics must be specific enough to score consistency yet flexible enough for judgment-based scenarios.
For example: “How would you score a model's response to a technical accounting question on deferred tax valuation allowances where facts are ambiguous?”
A task you may get
Review a short case of accounting work - perhaps a controls narrative or technical accounting treatment - and write a rubric that scores model responses on accuracy, reasoning quality, and real-world applicability.
How to prepare
- Review recent accounting standards updates and enforcement priorities from your domain (FASB, IFRS, PCAOB, SEC).
- Prepare examples of accounting judgments you made that required context beyond the technical rule.
- Document one audit or accounting situation where the textbook answer was wrong in practice.
- Understand how frontier AI models approach reasoning tasks and where they tend to cut corners.
The facts
- Pay
- $60–100/hr
- Hours
- Full time, 40 hours a week
- Where
- Hybrid · Bay Area, CA
- Open to
- USA
- Field
- Finance
- Posted
- 8/19/2026
- Places left
- 10
We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.