Senior Applied Scientist – AI Ranking & Content Quality

About the role

Microsoft’s Copilot Discover team is building the next generation of AI‑powered ranking and content‑quality systems that surface trustworthy, relevant content across Microsoft surfaces. As a Senior Applied Scientist you will lead the scientific direction of the ranking and quality stack, combining large language models, multimodal models, and large‑scale recommender systems to improve engagement, satisfaction, and trust.

Key responsibilities

• Lead content‑quality understanding at scale by designing models that assess credibility, usefulness, freshness, safety, and diversity across text, image, audio, and video.
• Champion safety and trust by partnering with policy teams to embed safety standards, create red‑team and adversarial safeguards, and reduce misinformation and toxicity.
• Scale end‑to‑end ML systems through collaboration with engineering on data contracts, feature stores, distributed training/inference, and automated rollout/rollback pipelines.
• Own evaluation and experimentation, defining offline metrics (e.g., Rejection Rate, ERR, Defect Rate) and online methodologies such as A/B tests, interleaving, counterfactual and bandit approaches.
• Provide technical leadership and mentorship to a high‑caliber science cohort, guiding problem framing, methodology selection, code quality, and knowledge sharing.

Required profile

• Senior‑level researcher with a proven track record in AI/ML, preferably in large‑scale ranking, recommendation, or content‑quality domains.
• Experience leading cross‑functional projects and influencing engineering, product, UX research, and policy teams.
• Strong publication record or equivalent impact in LLMs, multimodal learning, or recommender systems.
• Ability to design and ship end‑to‑end ML solutions that meet safety, reliability, and performance goals.

Required skills

• Large Language Models (LLMs)
• Multimodal Machine Learning
• Recommender Systems
• Distributed Training and Inference
• Feature Stores
• A/B Testing and Online Experimentation
• Counterfactual Learning
• Bandit Algorithms