About this role
Role Overview:
As an AI Quality Analyst, you will evaluate a new personalization feature for Gemini, focusing on how effectively the model utilizes information from your past conversations, Gmail, Google Search, and YouTube activity to enhance response relevance and helpfulness. This position requires a unique combination of creativity and analytical skills, as you will design prompts based on your personal experiences and assess the quality of the model''s personalized responses across various dimensions such as Grounding, Integration, and Helpfulness.
Key Responsibilities:
- Design and execute multi-turn conversational prompts (typically 1-5 turns) that leverage your personal information and experiences.
- Evaluate model responses based on your intent from the starting prompt, ensuring appropriate application of personalization.
- Analyze responses for Grounding issues, confirming that claims about you are supported by evidence rather than flawed inferences or hallucinations.
- Assess Integration quality to ensure personal data is seamlessly incorporated into responses without robotic "overnarrating".
- Conduct rigorous evaluations and stack-rank two model responses side-by-side (SxS) to identify which is more helpful, user-friendly, and enjoyable.
Qualifications:
- Dutch Proficiency: High-level reading and writing skills in Dutch, as this is the primary language for the project.
- Personal Account Usage: Willingness to use your primary personal Google account (not a testing account) and enable personal data sources for authentic assessments.
- Schedule Flexibility: Full-time availability in your local time zone is essential, as we operate a global, 24-hour team.
- Exceptional Analytical Thinking: Ability to evaluate nuanced and ambiguous AI responses, particularly in assessing personalization quality.
- Creative Prompt Engineering: Experience in crafting creative, multi-turn prompts based on personal context to thoroughly test the model''s capabilities.
- Strong Evaluation Acumen: Understanding of personalization concepts, including the ability to identify incorrect personalization, poor inferences, and forced connections.
- Meticulous Attention to Detail: Capability to review Side-by-Side (SxS) model responses and identify subtle differences in naturalness and overnarrating.
- Excellent Written Communication: Ability to write clear, concise, and structured rationales for model rankings, explicitly referencing specific turn numbers.
- Feedback: Skill in providing constructive feedback and detailed annotations.
- Communication: Strong communication and collaboration skills.
- Independence: Self-motivated and capable of working independently in a remote environment.
- Technical Setup: Desktop or laptop with a reliable internet connection.
Work Terms:
Contract position with full-time availability required.
Compensation:
$20 per hour.
Eligibility:
Must have proficiency in Dutch and be willing to use a personal Google account for evaluation tasks.