AI Evaluation Data Scientist

Role Number : 200618553-3337

Summary

The Health Sensing team builds outstanding technologies to support our users in living their healthiest, the happiest lives by providing them with objective, accurate, and timely information about their health and well-being. As part of the larger Sensor SW & Prototyping team, we take a multimodal approach using a variety of data types across HW platforms, such as camera, PPG, and natural languages, to build products to support our users in living their healthiest, happiest lives.

In this role, you will be at the forefront of developing and validating evaluation methodologies for Generative AI systems in health and wellbeing applications. You will design comprehensive human annotation frameworks, build automated evaluation tools, and conduct rigorous statistical analyses to ensure the reliability of both human and AI-based assessment systems. Your work will directly impact the quality and trustworthiness of customer-facing health products.

Description

In this role you will :

Design and analyze human evaluations of AI systems to create reliable annotation frameworks, and ensure validity and reliability of measurements of latent constructs

Develop and refine benchmarks and evaluation protocols, using statistical modeling, test theory, and task design to capture model performance across diverse contexts and user needs

Conduct statistical analysis of evaluation data to extract meaningful insights, identify systematic issues, and inform improvements to both models and evaluation processes

Analyze model behavior, identify weaknesses, and drive design decisions with failure analysis. Examples include, but not limited to : model experimentation, adversarial testing, counterfactual analysis, creating tools to assess model behavior and user impact

Collaborate with engineers to translate evaluation methods and analysis techniques into scalable, adaptable, and reliable solutions that can be reused across different features, use cases, and evaluation workflows

Work cross-functionally to apply methods to real-world applications with designers, clinical experts, and engineering teams across Hardware and Software

Independently run and analyze experiments for real improvements

Minimum Qualifications

BS and a minimum of 10 years relevant industry experience in a empirical field with emphasis on quantitative methodologies of human behavior, including HCI, Psychometrics, Quantitative or Experimental Psychology, Educational Measurement, Language Assessment, or a relevant field

Proficiency in Python and ability to write clean, performant code and collaborate using standard software development practices (e.g. Git)

Strong statistical analysis skills and experience in crafting experiments, validating data quality and model performance

Experience in building and extending data and inference pipelines to process large scale datasets

Preferred Qualifications

MS or PhD or equivalent experience in relevant fields

Real-world experience with LLM-based evaluation systems and human annotation and human evaluation methodologies

Experience in rigorous, evidence-based approaches to test development, e.g. quantitative and qualitative test design, reliability and validity analysis

Customer-focused mindset with experience or strong interest in building consumer digital health and wellness products

Strong communication skills and ability to work cross-functionally with technical and non-technical stakeholders

Apple is an equal opportunity employer that is committed to inclusion and diversity. We seek to promote equal opportunity for all applicants without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, Veteran status, or other legally protected characteristics. Learn more about your EEO rights as an applicant () .