Google Gemini
Defining the principles and operational rigor behind Gemini’s response quality.
The Challenge
As the business lead to the Design Leader of Model Content Quality, I was tasked with scaling the operational infrastructure for one of Google’s most ambitious AI initiatives. The primary challenge was ensuring consistent model response quality across a rapidly evolving landscape, particularly during a period of high public scrutiny regarding model accuracy and over-correction.
The Strategy
I implemented a multi-layered governance and alignment strategy designed to turn high-level policy into concrete product experiences. Key initiatives included:
Gemini App Principles: I co-led the foundational alignment and rollout of the core principles that define how Gemini interacts with users: following directions, adapting to needs, and safeguarding the experience.
Governance Framework: I operationalized a framework to translate "red line" policies into actionable principles for internal teams and vendors, ensuring a "safe but helpful pivot" on sensitive topics rather than simply punting.
Work-Wide Quality Rhythms: I established rigorous quality enforcement measures and "rituals," including team onboarding, training deliverables (Goldens and evals), and cross-functional alignment forums to ensure response quality remained the top priority.
Knowledge Sharing: I scaled a model for sharing content design learnings across the broader Google ecosystem, creating a shared "team brain" that bridged siloed expertise.
The Impact
By leading with strong partnerships, I moved the organization from tactical execution to strategic alignment. My work provided the structural backbone for the Gemini app's public rollout at Google I/O, ensuring that the teams and vendors responsible for model training had the clear, consistent instructions needed to drive high-quality, helpful AI responses.