CVML - Failure Analysis, Data Engineer
Santa Clara Valley (Cupertino), California, United States
Software and Services
iPhone is the most popular camera in the world. In our Camera & Photos organization (C&P), the precise integration of software and hardware has led to features like Memories and Portrait Mode, which deliver magical, user-focused experiences to 1 Billion+ devices. Within Camera and Photos, Computer Vision and Machine Learning group (CVML) is responsible for crafting machine learning solutions applied to Apple's image and video processing technologies. This includes technology for OS X and iOS system frameworks (e.g. face recognition and scene classification) as well as internal tools. The group combines research and development in a fast-paced environment. Our data team is responsible for building high quality datasets at scale for all features developed in CVML, and more generally in Camera and Photos. We are looking for a failure analysis engineer to challenge both models and datasets created by R&D teams across the organization, in order to make sure they deliver the quality expected from Apple products.
- You're aware of the challenges associated to building ML datasets and computer vision models: definition and coverage of target data distribution, bias
- You have creativity and a good sense of product, which enable you to have an intuition of how each model is expected to work (eg face detector), where a model could potentially fail (eg low light, pale/dark skin etc...), what data can be used to test failure patterns
- You have strong Python coding skills which enable you to manipulate data at scale & run ML models
- You come up with data-driven solutions when facing situations of trade-offs/decision under uncertainty
- You show strong communication skills and proactivity, and you’re comfortable interacting with multiple R&D teams
Our team works in close interaction with R&D, infrastructure and client teams, as well as with other groups and other functions across Apple (legal, privacy) and externally. In this position, you will partner with R&D DRIs (technical leads) and with clients of current and future models built by C&P to figure out how our models should perform. You will test these models at scale (with millions of images) to discover failure patterns from data. The role also includes the production of tools associated with failure analysis: testing framework, reporting, data visualization.
Education & Experience
Bachelors, Masters, or Ph.D. in Computer Science, Mathematics, Physics, or a related field (or equivalent practical experience)
- Interest in photography and experience in training CV/ML models that have shipped are a strong plus