LLM Safety Evaluator - Cantonese

We are hiring an LLM Safety Evaluator - Cantonese to support AI safety, adversarial evaluation, and language model quality testing for a leading technology client.

This role is focused on LLM red teaming and adversarial safety evaluation. The ideal candidate will have hands-on experience evaluating language model outputs, deliberately testing models to identify unsafe or harmful responses, and translating those findings into preventive frameworks, evaluation criteria, and actionable insights.

This is not a BI/dashboard-heavy role. Tableau and reporting experience are helpful, but the main priority is experience with LLM output evaluation, AI safety, red teaming, localization, linguistics, and Cantonese language expertise.

Key Responsibilities:
Evaluate the quality and safety of LLM/generative AI outputs.
Perform adversarial testing and red teaming to identify harmful, unsafe, biased, or policy-violating model responses.
Create and improve adversarial datasets, stress-test datasets, and model evaluation criteria.
Analyze model behavior across Cantonese language, culture, and market-specific contexts.
Translate qualitative findings into clear, actionable insights for model improvement.
Support evaluation frameworks that measure model safety, policy alignment, and response quality.
Use SQL and Python for data analysis where required, including pandas, NumPy, and Jupyter.
Collaborate with cross-functional teams working on AI safety, localization, linguistics, and model evaluation.

Required Qualifications:
Native Cantonese proficiency is mandatory.
Hands-on experience evaluating LLM or generative AI outputs for quality, safety, harmful content, bias, or policy violations.
Experience with LLM red teaming, adversarial prompt testing, AI safety evaluation, or model behavior analysis.
Background in localization, linguistics, trust and safety, content quality, language operations, or AI safety.
Ability to build or contribute to adversarial/stress-test datasets and evaluation frameworks.
Ability to translate qualitative and quantitative findings into actionable insights.
Working knowledge of SQL and/or Python for data analysis is preferred.

Preferred Qualifications:
Experience with Python libraries such as pandas, NumPy, or Jupyter.
Experience working with large-scale AI evaluation datasets.
Experience with safety policy evaluation, content moderation, annotation, linguistic QA, or localization QA.
Familiarity with Tableau or reporting dashboards is a plus, but not required.
Bachelor’s degree in Data Science, Linguistics, Computer Science, Data Analytics, or a related field is preferred, but not a dealbreaker.
Show more Show less