AIML – Sr Machine Learning Engineer, Data and ML Innovation Full-time

Summary
As part of Apple’s AI and Machine Learning org, we encourage and create groundbreaking technology for large-scale ML systems, computer vision, natural language processing, and multi-modal understanding. The Data and Machine Learning Innovation (DMLI) team is looking for a passionate Machine Learning Engineer to explore new methods, challenge existing metrics or protocols, and develop new insightful practices that will change how we understand data and overcome real-world ML challenges. Are you excited to work on some of the most ambitious technical challenges in the field? Your role will involve collaborating closely with machine learning researchers, engineers, and data scientists. Together, we will spearhead groundbreaking research initiatives and develop transformative products designed to build a significant impact for billions of users worldwide.

Description
As a Machine Learning (ML) Engineer, you will be entrusted with the critical role of innovating and applying innovative research in ML to tackle complex data problems. The solutions you develop will significantly impact future Apple products and the broader ML development ecosystem. You will work with a multidisciplinary team to actively participate in the data-model co-design and co-development practice. Your responsibilities will extend to the design and development of a comprehensive data curation framework. You will also build robust model evaluation pipelines, integral to the continuous improvement and assessment of ML models. Additionally, your role will entail an in-depth analysis of collected data to underscore its influence on model performance. Furthermore, you will have the opportunity to showcase your groundbreaking research work by publishing and presenting at premier academic venues. Your work may span a variety of topics, including but not limited to:

– Designing and implementing semi-supervised, self-supervised representation learning techniques for growing the power of both limited labeled data and large-scale unlabeled data.

– Developing evaluation protocols centered on the end-to-end user experience, with a focus on anticipating potential failure modes, edge cases, and anomalies.

– Employing data selection techniques such as novelty detection, active learning, and core-set selection for diverse data types like images, 3D models, natural language, and audio.

– Uncovering patterns in data, setting performance targets, and using modern statistical and ML-based methods to model data distributions. This will aid in reducing redundancy and addressing out-of-distribution samples.