Senior Machine Learning Engineer

Microsoft is hiring a Senior Machine Learning Engineer in Cambridge, MA, onsite, to help advance generative models for multimodal scientific discovery. The role emphasizes designing, training, and optimizing training workflows across the model lifecycle, while partnering closely with researchers and domain experts to translate new research into practical outcomes.

In this position, you will lead large-scale training work for language and multimodal systems, improve throughput and efficiency, and provide technical direction across the team and organization. This opportunity sits within Microsoft Research New England, a group of full-time researchers, postdocs, data scientists, engineers, and program managers spanning multiple disciplines.

Key responsibilities

• Lead the training of large language and multimodal models by developing, managing, and debugging training workflows
• Scale and optimize training runs by identifying efficiency opportunities and implementing improvements
• Mentor and provide technical leadership within the team and across the organization in your areas of expertise
• Explore and advise on strategic direction and technical approaches for projects across the team
• Collaborate with researchers, domain experts, and cross-disciplinary teams to ensure techniques transfer effectively from research to impact

Requirements

• Bachelor’s Degree in Computer Science (or related technical field) AND 4+ years of technical engineering experience in ML-related work, or an equivalent combination including a Doctorate in Computer Science (or related technical field) with 4+ years of ML-related engineering experience
• Hands-on technical experience in at least one of the following: distributed training of large language or multimodal models (pretraining, midtraining, or post-training), optimizing training workflows, or model distillation
• Deep experience with PyTorch or Jax
• Experience designing, optimizing, and monitoring generative modeling training workflows
• Familiarity with profiling training and/or inference code, identifying performance bottlenecks, and mitigating them
• Comfort with one or more of: refining data mixes, midtraining, post-training, data/model parallelism techniques, optimizing training runs, or model distillation
• Demonstrated communication skills, including the ability to present complex technical concepts to a broad audience

Technologies

• PyTorch
• Jax

Role overview

This senior role focuses on building capability in generative models for multimodal scientific discovery. There is no single universal approach, so progress depends on ongoing investment across the training lifecycle, including method development, recipe refinement, scaling and throughput optimization, and model-harness co-design.

Compensation: USD 119,800 - 261,000 per year.