Conduct deep-dive loss analyses, data exploration, and prototyping to identify high-leverage opportunities and quality headroom across Factuality, Helpfulness, and Freshness. Design and implement scalable systems, ranking formulations, ML scorers, and prompt/signal augmentations to improve input ranking (Octopus), end-to-end context selection for AIO and AIM. Evaluate the impact of context and ranking changes on both input-level metrics and downstream AI answer quality, driving changes from offline autorater evaluations through Live Experiments to full production launch.
Work across a wide breadth of systems spanning Google’s iconic core Search retrieval and ranking stack (QRS, UberSearch, Tangram, Web Ranking) through modern AI answer quality, ML modeling, serving infrastructure such as MaRS and Gemini/Magi orchestration and evaluation pipelines. Partner closely with Web Ranking, first-party Verticals, Serving Infrastructure, and Large Language Model (LLM) Modeling teams to align Search signals.
Job description provided by Google.
View the original posting