Microsoft 365 aims to empower users to work, create, and communicate naturally through digital pen, voice, visual, and text-based interfaces by integrating deep technical expertise with platform-scale innovation. This role advances both the foundational AI platform and natural user interaction capabilities, shaping the future of productivity for hundreds of millions of users.
Requirements
- 2+ years of experience in coding in Python and modern ML frameworks such as PyTorch (or equivalent).
- 1+ year of experience training and/or fine-tuning large language, vision-language, or diffusion models (e.g., GPT, DeepSeek-R1, Stable Diffusion, Phi-V, Qwen, Llama) + OR Prompt Engineering
- Hands-on experience orchestrating LLM-based agent pipelines - prompt design, tool invocation, and context grounding - for productivity scenarios, alongside other advanced ML techniques.
- Proven expertise and background in designing and implementing AI models that serve millions of users in scaled production environments.
- Demonstrated success moving applied research into shipped product features, with clear ownership of the end-to-end ML lifecycle—data collection, cleaning, modeling, evaluation, and continuous improvement.
- Experience in designing, evaluating, and iterating on multimodal or agent-based workflows for real-world productivity applications.
- Deep understanding of transformer-based architectures and hands-on experience with model optimization techniques (quantization, pruning, distillation, adapter tuning).
Responsibilities
- Responsible for end-to-end development, deployment, and continuous improvement of advanced multimodal ML/LLM systems—including dataset curation, labeling strategy, model architecture, scalable training, and rigorous evaluation (offline and online).
- Research, design, and implement on-device and cloud-based AI models, optimizing for low latency, privacy, and resilience in both connected and constrained environments.
- Drive complex orchestration of Copilot workflows and agent pipelines, blending prompt engineering, tool invocation, and context grounding for productivity scenarios.
- Collaborate with engineering and product teams to define success metrics, design and execute experiments (e.g., A/B testing), monitor user signals, and iterate on deployed features.
- Advance the state-of-the-art through applied research, prototyping new concepts in ink, voice, vision, and LLM-based agents, and translating breakthroughs into scalable, user-centric product features.
- Apply Microsoft’s Security and Responsible AI principles throughout the ML lifecycle, ensuring robust data governance, fairness, transparency, and ethical AI practices.
- Stay current with research and industry trends, identifying opportunities to integrate emerging technologies and methods into the Office AI ecosystem.
Other
- The role is a leadership role.
- Close partnerships with engineering, product, Microsoft Research, and cross-functional stakeholders are central to delivering seamless multi-modal and AI-driven capabilities at scale.
- The team values inclusivity, growth mindset, and a culture that fosters innovation and impact.
- Mentor and influence peers, set technical direction, foster a collaborative and inclusive culture, and contribute to team talent development.
- Customer-obsessed mentality and passion for translating research breakthroughs into meaningful product impact.
