RAG Retrieval Error Analysis Specialist

Globik AI; United States; À distance; USD 17 000 - 25 000; Plein temps

Résumé du poste

Globik AI is seeking RAG Retrieval Error Analysis Specialists to identify and categorize the reasons retrieval systems fail to return useful information.

You will analyze failure examples—irrelevance, missing evidence, duplicates, poor context, and source mismatches—then record structured categories and patterns for evaluation and engineering teams.

Qualifications

• Review failed RAG retrieval examples.
• Identify failure causes: irrelevance, missing information, duplication, poor matching, or other issues.
• Compare retrieved content with user queries and expected evidence.
• Document findings for evaluation and engineering teams.

Responsabilités

• Review failed RAG retrieval examples.
• Identify whether the failure is caused by irrelevance, missing information, duplication, poor matching, or another project-defined issue.
• Compare retrieved content with the user's query and expected evidence.
• Categorize retrieval failure types consistently.
• Identify recurring failure patterns and edge cases.
• Document findings for evaluation and engineering teams.
• Maintain accurate error labels across large datasets.
• Participate in calibration and quality reviews.

Description du poste

Globik AI is seeking RAG Retrieval Error Analysis Specialists to identify and categorize the reasons retrieval systems fail to return useful information.

You will analyze retrieval failures such as irrelevant results, missing evidence, duplicate passages, poor context selection, and incorrect source matching, then record structured failure categories.

Responsibilities

• Review failed RAG retrieval examples.
• Identify whether the failure is caused by irrelevance, missing information, duplication, poor matching, or another project-defined issue.
• Compare retrieved content with the user's query and expected evidence.
• Categorize retrieval failure types consistently.
• Identify recurring failure patterns and edge cases.
• Document findings for evaluation and engineering teams.
• Maintain accurate error labels across large datasets.
• Participate in calibration and quality reviews.

Pay

$12-$18/task