LLM Evaluator (Model Response Analyst)

Odixcity Consulting; South Africa; Remoto; ZAR 360.000 - 540.000; Tempo pieno

Descrizione del lavoro

Odixcity Consulting seeks a detail-oriented LLM Evaluator to assess and improve large language models. You will analyze model outputs for factuality, coherence, safety, and alignment with guidelines, providing actionable feedback to ML teams.

You will review multiple responses, justify preferences, and help calibrate scoring with global evaluators, contributing to a living library of model behaviors and failure modes.

Competenze

• Minimum 2 years of professional experience in NLP/AI-related fields.
• Bachelor’s degree in CS or related field.
• Ability to explain why text is “good” or “bad” based on logic, tone, factuality and adherence to guidelines.
• Experience with RLHF data collection.
• Experience coordinating evaluation teams and IAA calibration.

Mansioni

• Evaluate and rank model outputs using rubrics covering factuality, coherence, safety, instruction-following, and creativity.
• Review multiple responses to a prompt and justify preferred output with clear reasoning.
• Provide feedback to ML teams on recurring failures observed during evaluation.
• Craft prompts to challenge models and reveal safety vulnerabilities.
• Collaborate to update evaluation guidelines for edge cases and ambiguities.
• Participate in cross-checking sessions to calibrate scoring across evaluators.
• Analyze root causes of errors beyond surface scores.
• Flag novel model behaviors to researchers for library of outputs and failure modes.

Conoscenze

Computational Linguistics

Data Analysis

Technical Writing

Quality Assurance (NLP/AI)

Cognitive Science

Prompting & Prompt Design

RLHF Data Collection

Formazione

Bachelor’s degree in Computer Science or related field

Descrizione del lavoro

Job Title: LLM Evaluator (Model Response Analyst)

Location: Remote (Worldwide)

Job Summary: We are seeking a detail-oriented and analytical LLM Evaluator to assess, analyze, and improve the performance of large language models (LLMs). In this role, you will evaluate AI-generated content for accuracy, coherence, factual reliability, bias, safety, and alignment with defined guidelines.

Responsibilities

• Evaluate and rank model-generated text based on complex rubrics covering dimensions such as factuality, coherence, safety, instruction- following, and creativity.
• Review multiple model responses to the same prompt and determine which output a human would prefer, providing justifications for your choices.
• Provide clear, concise feedback to the modeling and training teams regarding recurring failure models observed during evaluation sessions.
• Attempt to "break" the model by crafting prompts designed to elicit biased, harmful, or insecure outputs to help patch safety vulnerabilities.
• Collaborate with the quality assurance team to suggest improvements to evaluation guidelines when you encounter ambiguous or unclassifiable edge cases.
• Participate in regular "cross-checking" sessions with other evaluators to calibrate scoring standards and ensure inter-rater reliability across the global team.
• When a model underperforms, dig deeper than the surface score to hypothesize "why" the model made a specific error (e.g., training data vs. prompt misinterpretation).
• Identify and flag novel or unexpected model behaviors to the research team, contributing to a living library of unique model outputs and failure modes.

Requirements

• Minimum of 2 years of professional experience in a relevant field such as; Computational Linguistics, Data Analysis, Technical Writing, Quality Assurance (specifically for NLP/AI), or cognitive science.
• Bachelor’s degree in Computer Science, or a relating field.
• Deep understanding of how-to craft prompts to elicit specific behaviors and test model limits.
• Ability to look at a text output and explain "why" it is "good" or "bad" based on logic, tone, factuality, and instruction adherence.
• Experience working with Reinforcement Learning from Human Feedback (RLHF) data collection.
• Proven experience monitoring and improving consistency among evaluation teams. Ability to analyze IAA scores and conduct calibration sessions to align judgement.
• Experience sourcing, cleaning, and annotating datasets specifically for the fine-tuning or evaluating LLMs. Understanding of data distribution and its impact on model performance.
• Familiarity with A/B testing concepts applied to AI. Ability to help design experiments to test if a new model version is truly "better" than the previous one.