About the job
As a Group Product Manager for the Gemini App, you will not simply oversee app deployment; you will be instrumental in establishing the standards for "Response Quality" that will shape the AI experience within Google’s vast ecosystem of billions of users. Unlike traditional metrics, our team excels in understanding the subtle nuances of key user segments, transforming Gemini into a reliable, context-sensitive assistant. This is an exceptional opportunity to spearhead the advancement of personalized, high-utility experiences that will set a global standard for human-AI interaction.
About Us
At Google DeepMind, we believe that Artificial Intelligence can be one of humanity’s greatest assets. Our diverse team of scientists, engineers, and machine learning specialists collaborates to push the boundaries of AI technology for the benefit of society and scientific advancement. We prioritize safety and ethics in our work, ensuring that our innovations serve the public good while tackling some of the world’s most pressing challenges.
Your Role
You will be the guiding force behind the "intelligence and soul" of the Gemini experience. Your responsibilities include establishing evaluation frameworks for response quality and leading collaborative initiatives to optimize the model's outputs for critical user journeys. You will operate at the vital intersection of DeepMind research and the Gemini App, significantly impacting model behavior by translating intricate human needs into actionable data signals that inform future training cycles.
Key Responsibilities:
- Define the Quality Bar: Establish and uphold North Star metrics for Response Quality, articulating what "helpfulness" means across various formats including text, code, and multimodal channels.
- Cohort-Specific Optimization: Assess the needs of critical user segments (e.g., power users vs. casual seekers), identify performance gaps, and prioritize model adjustments accordingly.
- Champion User Journey Excellence: Ensure that the model delivers a seamless, accurate, and delightful experience for key user interactions.
- Implicit Alignment & Tuning: Collaborate with Research and Data Operations to refine model outputs through Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), converting qualitative insights into structured data strategies.
- Evaluation Strategy: Develop and scale advanced human-in-the-loop and "LLM-as-a-judge" evaluation frameworks to assess model performance during Side-by-Side (SxS) testing.