Apple’s AIML residency is a year-long program inviting experts in various fields to apply their own domain expertise to innovate and build revolutionary machine learning and AI-based products and experiences. As AI-based solutions spread across disciplines, the need for domain experts to understand machine learning and apply their expertise in ML settings grows.
Residents will have the opportunity to attend ML and AI courses, learn from an Apple mentor closely involved in their program, collaborate with fellow residents, gain hands-on experience working on high-impact projects, publish in premier academic conferences, and partner with Apple teams across hardware, software, and services.
Our team is part of Apple's Machine Learning Research organization. We conduct fundamental research in two areas: 1) Efficient machine learning, focusing on optimizing models and algorithms for high performance while minimizing resource usage by improving data efficiency, reducing training time, and lowering inference compute and memory demands; and 2) controllable generation, sought at improving the controllability and capabilities of multi-modal generative models. This enables techniques such as creating and augmenting virtual worlds with sophisticated controls, and generating synthetic data for training downstream models. A few recent representative research works from our team include:
- “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models (https://machinelearning.apple.com/research/gsm-symbolic)”, arXiv 2024
- “Probabilistic Speech-Driven 3D Facial Motion Synthesis: New Benchmarks, Methods, and Applications (https://arxiv.org/abs/2311.18168)”, CVPR 2024
- “MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training (https://arxiv.org/abs/2311.17049)”, CVPR 2024
- “HUGS: Human Gaussian Splats (https://machinelearning.apple.com/research/hugs)”, CVPR 2024
- “Corpus Synthesis for Zero-shot ASR domain Adaptation using Large Language Models (https://arxiv.org/abs/2309.10707)”, ICASSP 2024
- “Knowledge Transfer from Vision Foundation Models for Efficient Training of Small Task-specific Models (https://arxiv.org/abs/2311.18237)”, ICML 2024
- “Tic-clip: Continual training of clip models (https://machinelearning.apple.com/research/tic-clip-v2)”, ICLR 2024
Our Resident will focus on research at the intersection of multi-modal agents (vision, language, audio), foundation models and data-centric learning methods.