AI training and LLM evaluation jobs
Accountant (CPA/CA)
Apply on micro1 ↗View source posting ↗
Tax professionals will translate sales-and-use-tax law into precise evaluation rubrics for AI outputs. They must read statutes at subsection level, identify conditions, thresholds, exceptions, and logical relationships, then score formal translations against independently written criteria. Failed criteria require self-contained test scenarios with explicit facts and expected outcomes. The source requires an active CPA, Chartered Accountant qualification, or US tax-attorney experience; state and local tax expertise is preferred. This role rewards statutory interpretation and legal logic rather than routine return preparation.
Responsibilities
- Read and interpret sales and use tax statutes and regulations across US state jurisdictions, and occasionally federal tax provisions, at the level of individual subsections.
- Author rubrics: break a section of law into a checklist of discrete requirements the AI's translation must satisfy, grounded strictly in the statute text and written independently of the AI's output.
- Assign severity to each requirement (critical vs. secondary vs. minor) based on genuine statutory weight, not defaulting everything to critical.
- Review the AI's formal translation of a statute and score it pass/fail against your rubric, checking that conditions, thresholds, AND/OR logic, and exceptions are captured accurately.
- Write clear, self-contained test scenarios (including edge cases) for failed criteria, with explicit facts, dates, amounts, and expected outcomes.
- Identify and flag discrepancies, statutory exceptions, and genuine ambiguities in the law.
- Work within a digital review platform, submitting structured written evaluations.
Requirements
- Active CPA license, Chartered Accountant designation, or experience as a US tax attorney.
- Sales and use tax or broader state and local tax statutory experience strongly preferred.
- Proven ability to read and interpret tax statutes and regulations, not just prepare returns.
- Comfort working with legal logic: conditions, thresholds, AND/OR structures, exceptions, and defined terms.
- Strong written communication with a focus on precision. Another reviewer should reach the same conclusion from your criteria.
- High attention to detail, especially around "except for" clauses and edge cases.
- Comfort providing structured feedback through a digital review interface.
Eligibility
Bangladesh · Hong Kong · India · Indonesia · Japan · Kazakhstan · Kyrgyzstan · Malaysia · Pakistan · Philippines · Singapore · Sri Lanka · Taiwan · Thailand · Uzbekistan · Vietnam · Austria · Belarus · Belgium · Denmark · France · Germany · Greece · Italy · Netherlands · Portugal · Russia · Spain · Switzerland · United Kingdom · Argentina · Brazil · Chile · Colombia · Mexico · Peru · Algeria · Bahrain · Egypt · Iraq · Jordan · Kuwait · Lebanon · Libya · Morocco · Oman · Palestine · Qatar · Saudi Arabia · Tunisia · United Arab Emirates · United States · Canada · Nigeria · Kenya · South Africa · Ghana · Ethiopia