Machine Learning Research Engineer - Fitness Intelligence, Watch Software

Summary

magine what you could do here at Apple. Innovation runs through everything we do, from amazing technology to industry-leading environmental efforts, and the diversity of people and ideas makes it possible. Join Apple, and help us leave the world better than we found it. The Apple Watch Software team is looking for an exceptional machine learning engineer to help ship the intelligence behind Apple’s fitness experiences. You’ll take promising model capabilities from first prototype all the way to features on Apple Watch and iPhone. That means building the datasets, evaluations, prompts and model adaptations that get a capability to shipping quality, and writing the production client code that delivers it to users. You’ll work closely with the other fitness teams to bring unified intelligence capabilities to features across the platform, and you’ll help set the technical direction for a new and fast-moving domain. This role requires a passion for iteration, a high bar for quality, and a drive to see ideas all the way into shipping software. We value communication skills, integrity and creativity, and aspire to build diverse, balanced and dynamic teams.

Minimum Qualifications

8+ years of industry experience, or a PhD in machine learning or a related field plus 5+ years of industry experience. A track record of shipping user-facing features built on machine learning, carried through the iteration needed to reach production quality. Deep, hands-on experience with large language models, including prompting, fine-tuning or adaptation. Strong command of evaluation methodology: dataset design, metric development, human and automated evaluation, and understanding the limits of each. Strong software engineering skills in a compiled language and/or proficiency in Python and the modern ML stack (e.g. PyTorch or JAX). Strong product sense and the ability to balance technical ambition, feasibility and user experience. Excellent communication skills, including the ability to explain technical tradeoffs clearly to engineering and product partners.

Preferred Qualifications

Experience shipping production Swift, including API design and performance and memory budgets. Experience deploying models on device, including quantization, compression or latency optimization. Experience building datasets and measurement systems, and using them to make quality tradeoffs defensible. A track record of influencing outcomes across organizational boundaries and delivering work that other teams’ features depend on. Experience with time-series, sensor or multimodal data, especially from wearables. Background in health, fitness or another domain where user trust and data sensitivity shape the design. Familiarity with privacy-preserving machine learning techniques. Publications at leading ML venues, or equivalent impact in industry.

Description

n this role you’ll be a senior technical leader in a small, deeply technical team working where applied machine learning meets fitness. You’ll write production Swift that ships in the OS and runs both on device and in server-side environments. You’ll also build the evaluations, datasets and model improvements that make those features good. You’ll work closely with the Activity, Workout and Fitness+ engineering teams, as well as with Siri and platform intelligence teams across Apple. You’ll also help build the tools and harnesses that let engineers across the organization develop and evaluate intelligence-backed features with confidence.

Responsibilities

Partner with feature-owning teams to define what the intelligence layer must deliver, then deliver it, co-owning outcomes rather than surfaces. Write production client software in Swift, including APIs, on-device inference integration and feature logic, within real performance, power and memory budgets. Own the iteration loop to shipping quality: prompt development, tool and function definitions, scenario construction, model adaptation and how they work together. Design golden datasets, metrics and evaluation pipelines that give the organization a trustworthy answer to whether a capability is good enough to ship. Bridge offline evaluation and production inference, so that what we measure matches what users experience. Drive the techniques and infrastructure that let features scale past their first release, including additional languages and regions, broader feature awareness and richer interactivity. Track the state of the art in machine learning, prototype what looks promising, and make deliberate, evidence-based calls about which new techniques are ready for shipping experiences and which are not. Raise the technical bar for the team through design reviews, mentorship and clear written communication. Design for user privacy from the first sketch, because fitness and health data deserves nothing less.