About the job
The team you will join is responsible for creating the technologies that power new, innovative product features for Apple Watch, like DoubleTap, AssistiveTouch, Handwashing, and Raise to Speak. We are highly collaborative and partner with a variety of research and product teams across Apple to explore novel experiences and ship features. We are looking for a versatile Machine Learning Software Engineer who is passionate about developing innovative, ML-driven product features that push the boundaries of sensing and human-computer interaction, and who can work across disciplines — from training advanced models to rapid on-device prototyping and full-scale productization.
Responsibilities
- Develop and optimize ML algorithms leveraging multimodal sensor data — like motion and audio — to detect user activities and contextual situations that enhance our understanding of real-world behavior
- Integrate and deploy ML models on-device, building power-efficient frameworks that encapsulate models, interface seamlessly with sensors, and communicate effectively with UX layers
- Drive innovation from concept to deployment, ensuring promising research ideas evolve into high-impact, user-facing features
- Design and implement tools, analytics, and processes to perform in-depth, hands-on analysis for validating and quantifying algorithm performance both offline and on-device
Qualifications
Minimum
M.S. or Ph.D. in Machine Learning, Computer Science, or a related field
Solid knowledge of machine learning methods, statistical analysis, and predictive modeling using time-series data
Strong Python skills with experience writing production-quality code and working with deep learning frameworks such as PyTorch or TensorFlow
Experience with Swift or Objective-C and developing on Apple platforms
Excellent communication and collaboration skills, with ability to work independently or in small teams
Preferred
Proficient in the full ML development cycle: data collection, model training and optimization, defining metrics, evaluation, performing failure analysis, and model deployment to resource constrained devices