SWE-Robustness Analysis ML Engineer, SIML

Zurich, Zurich, Switzerland
Software and Services

Summary

Posted:
Weekly Hours: 40
Role Number:200361419
The promotion of inclusion and fairness is a priority for Apple. It applies to our products and features, including those which are powered by machine learning models! The System Intelligence and Machine Learning (SIML) group is responsible for crafting machine learning solutions which transform the way millions of people capture, discover and share the most special moments of their lives, on all Apple platforms (macOS, iOS, tvOS, watchOS). Examples include face detection, scene classification, OCR, handwriting recognition... The group combines research and development in a dynamic and engaging environment. Within SIML, our data team assesses the robustness of features powered by ML, in addition to designing and building high quality datasets at scale. At the heart of machine learning, data defines how Apple features and products operate and what is the final user experience that will impact millions of our customers. This is an exciting time to join us: grow fast, and have a positive impact on multiple key features on your first day at Apple!

Key Qualifications

  • You have experience in training machine learning models (preferably text or computer vision)
  • You're aware of the challenges associated to building ML datasets and machine learning models (eg face detector): definition and coverage of target data distribution, potential biases, potential failure modes (eg low light, pale/dark skin etc...)
  • You have strong Python coding skills which enable you to manipulate data at scale & to run ML models
  • You have creativity and a good sense of product, which enable you to create tools/frameworks that are impactful and convenient to use by different types of end users
  • You’re hard-working, pragmatic and result-driven
  • You show strong communication skills and proactivity

Description

Our data team historically focused on data acquisition, data science, and data annotation. Each year, we power dozens of features and work closely with ML teams across the entire company. Apple’s commitment to deliver incredible experiences to a global and diverse set of users, in full respect of their privacy, has recently led to the development of a new function, Robustness Analysis (RA). RA IS A DATA DRIVEN INITIATIVE TO: - monitor model performance on relevant axes - surface, measure and mitigate ML failure modes, in order to improve overall user experience and reduce risks, with specific attention given to inclusion and fairness In this position, you will join a team of pioneers who contribute to the maturation of a robustness analysis function across SWE, by leading RA operations for key Apple features. This includes: - define product-centered axes of analysis (eg lighting conditions for images) and associated classes (eg low light < 100 ISO, backlit, ...) relevant to target feature, in collaboration with model DRI and feature DRI - identify other potential failure modes based on a data driven approach - when applicable, benchmark model using targeted public datasets - characterize potential biases in training set along chosen axes - request data collection efforts and/or implement "smart" pipelines based on advanced ML technology and humans in the loop to create test sets covering the various axes of investigation - report progress and issues found in technical meetings and sponsor meetings - suggest mitigation options (data and/or model) and lead mitigation experiments, when issues are found - contribute to the emergence of standardized RA processes and to the maturation of specs to drive the creation/improvement of RA tools In this position, you will interact with product teams and with R&D DRIs.

Education & Experience

Bachelors or Masters in Computer Science, Mathematics, Physics, or a related field (or equivalent practical experience)

Additional Requirements