The client, a leading global player in the AI space, is looking to train machine learning models to understand human language more accurately, and needs high-quality, structured feedback (annotations) to achieve this goal.
Requirements
- Proven experience in data entry, content moderation, data annotation, or linguistics.
- Excellent command of the English language, including understanding slang, idioms, and cultural context.
- Ability to follow complex, detailed guidelines and flowcharts without deviation.
- Strong research skills to quickly verify facts and information during the annotation process.
Responsibilities
- Evaluate pairs of AI-generated responses and rank them based on specific criteria such as helpfulness, honesty, and harmlessness (RLHF - Reinforcement Learning from Human Feedback).
- Annotate text data by tagging it with relevant categories (e.g., sentiment analysis, topic classification, safety violations).
- Write and refine prompts to test specific model behaviors, ensuring the data covers a wide range of conversational scenarios.
- Verify the factual accuracy of AI outputs by cross-referencing with reliable sources and labeling inaccuracies.
- Identify and report patterns of model failure, such as repetitive phrasing, biases, or failure to follow constraints.
Other
- Ability to work independently in a remote environment, managing time across multiple projects.
- Comfortable working 5-10 hours per week, with the option to work up to 40 hours.
- Must be legally authorised to work in the country where you are based (United States, United Kingdom, Canada, Ireland, Australia & New Zealand)
- Degree requirements not specified, but experience in relevant fields is required
- No travel requirements mentioned
