Complex Data Reconciliation Benchmark Author

ixolabs.ai; France; Hybride; EUR 30 000 - 42 000; Plein temps

Résumé du poste

IXOCreate seeks to design evaluation tasks that test whether an AI system handles demanding data-entry and reconciliation work correctly. You will assemble data from diverse sources, create credible errors, and specify exact corrections to reach a valid final state.

Deliverables include a complete benchmark package with inputs, task instructions, expected outcomes and a detailed scoring rubric. Prior AI experience is not required; strong documentation and remote collaboration are essential.

Qualifications

• Must have strong data entry and QA experience in regulated or audit-sensitive contexts.
• Ability to document decisions and explain changes clearly.
• Excellent written English with precise attention to file-level details.

Responsabilités

• Assemble complex datasets from CSV files, PDFs, spreadsheets and technical documents.
• Introduce credible test cases such as malformed records, missing values, inconsistent formats and silent truncation.
• Specify the correct final state and precise corrections needed to reach it.
• Write a detailed scoring rubric with at least 35 criteria for accuracy and error detection.
• Reproduce care expected in regulated processes without private data.
• Revise task packages after review and explain the rationale for changes.

Connaissances

Data entry

Quality assurance

Validation

Attention to detail

Written English

Description du poste

Your work at IXOCreate evaluation tasks that test whether an AI system can handle demanding data-entry and reconciliation work correctly. The challenge is to design realistic inputs and an unambiguous correct result, including errors that a superficial review could miss.

Responsibilities

• Assemble complex datasets from CSV files, PDFs, spreadsheets and technical documents.
• Introduce credible cases such as malformed records, missing values, inconsistent formats and silent truncation.
• Specify the correct final state and the precise corrections or reconciliations needed to reach it.
• Write a detailed grading rubric containing at least 35 criteria, covering both accurate corrections and complete error detection.
• Reproduce the care expected in regulated or audit-sensitive processes without relying on private operational records.
• Revise task packages after review and explain the reasoning behind each change.

Preferred background

Relevant experience includes data entry, quality assurance and validation in fields such as claims processing, back-office operations or legal administration. Strong candidates can explain how they measured accuracy, detected subtle inconsistencies and documented outcomes in work where errors mattered.

Clear written English, close attention to file-level detail and an organized approach are important. You should be comfortable communicating independently in a remote team and turning practical knowledge into explicit evaluation instructions. Prior AI experience is not required.

Deliverables and terms

Each deliverable is a complete benchmark package: the input files, task instructions, expected final state and a sufficiently detailed scoring rubric. IXO will agree the number of tasks, any minimum submission schedule and the review process before work starts.

Per completed task; fee agreed before work. The amount will be agreed in USD for the specified deliverable. An hourly estimate is not a promised rate, and no hourly conversion is advertised. Use synthetic, public or authorized data in the task materials.