Join Microsoft Cloud Operations and Innovation (CO+I) as a Principal Applied Scientist and shape the future of global data center operations. This role is central to integrating cutting-edge, multimodal agentic AI systems into core operational workflows, driving efficiency at an unprecedented scale. You'll leverage intelligent AI agents to empower every person and organization, aligning with Microsoft's mission to foster global achievement.

As a key member of the Data Center Applied AI team, you will translate state-of-the-art AI research into production-grade engineering solutions. Focus on building reliable, secure, and scalable agent architectures. Collaborate closely with business and engineering teams to implement agentic capabilities across diverse areas like workflow automation, information retrieval, retrieval-augmented generation (RAG), tool usage, and multi-agent orchestration.

Advance Applied AI research by applying deep expertise in Generative AI, deep learning, NLP, and multimodal models to create high-impact, production-ready solutions. Design and run experiments to enhance agent planning, memory, grounding, reasoning, and long-horizon task completion. Perform lightweight fine-tuning on multimodal and large multimodal language models to boost entity recognition, reasoning accuracy, and response quality.

Build and deploy scalable agentic AI systems by architecting end-to-end workflows that handle user intent, tool invocation, and error recovery. Rapidly prototype solutions and work with engineering teams for production deployment, including debugging live systems and developing AIOps workflows. Use data and telemetry to identify AI quality gaps, generate insights, and deliver proofs of concept that apply research innovations to real-world operational challenges. Design robust multi-step reasoning and tool-use strategies with strong safety and reliability guardrails.

Own evaluation, metrics, and iteration loops by defining task-level evaluation metrics for agentic behavior, such as success rate, tool-call accuracy, and hallucination rate. Build and maintain offline and online evaluation pipelines, including golden datasets, scenario simulators, and human-in-the-loop frameworks. Collaborate across disciplines by partnering with business and engineering stakeholders to translate requirements into technical specifications. Lead design reviews and mentor other applied scientists and engineers, elevating the standard for applied AI execution and production quality.

A PhD in Computer Science, Machine Learning, AI, NLP, or a related field is required, or an MS with substantial experience in delivering production-grade AI systems. You should have over 10 years of hands-on experience applying NLP and/or LLMs to real-world production challenges. A strong foundation in Large Language Models (LLMs), Generative AI, deep learning, and modern NLP is essential, coupled with proven experience building production-grade LLM systems, including prompt engineering, RAG, and multi-step reasoning pipelines.

Solid expertise in chunking strategies and embedding selection, understanding tradeoffs in retrieval quality, context size, and downstream task performance, is crucial. Experience designing and operating vector databases, including indexing strategies, similarity metrics, and retrieval optimization, is also required. You must have experience supporting production operations (AIOps/MLOps), covering monitoring, logging, quality metrics, and debugging live systems.

Proficiency in Python and experience with modern ML frameworks like PyTorch or TensorFlow, along with model serving/inference, is necessary. Strong applied experimentation skills, including defining metrics, analyzing results, and iterating based on data, are vital. The ability to collaborate effectively with engineering, product, and business stakeholders is key to success in this role.