Staff Applied Scientist, UI Control Models

DESCRIPTION

In this role you’ll drive the design, development and delivery of large language models that power the core of Siri. You will lead research and engineering efforts in model post-training techniques that create powerful UI control, tool calling and compelling conversational models that run within the constraints of mobile hardware. You’ll collaborate closely with researchers, software engineers and product teams to move innovations from prototype to production.

This includes identifying and applying post-training techniques, advancing model efficiency and contributing to the end-to-end model lifecycle process. This position emphasizes technical leadership through action, hands-on innovation and impact. You will be responsible for setting technical directions for the team. Our team values rapid iteration and delivering impactful results. The ideal candidate will have a passion for applied research, strong modeling expertise, good software engineering skills and a desire to get things done.

Masters or PhD in Computer Science, Machine Learning Engineering or equivalent professional experience 7+ years of hands-on industry experience and a proven track record of shipping ML-powered products or features, ideally involving LLMs. Extensive experience with the ML development loop, with expertise in model architectures, data curation, training policies and evaluation. Deep expertise in LLM/VLM post-training methods, including supervised fine-tuning, and reinforcement learning.

Proficiency in Python and one of the deep learning frameworks (such as PyTorch, JAX, or Tensor Flow). Excellent communication: strong interpersonal, verbal and written skills.

Experience developing AI agents capable of complex reasoning, planning, tool use and UI control in diverse environments. Experience developing small LLMs for constrained or edge devices. Experience building thoughtful benchmarks and environments to evaluate UI control agents. Deep understanding of model efficiency and deployment trade-offs in production environments. Experience with multi-modal foundation models. Experience developing ML systems for natural language understanding or digital assistants.