Generative AI Applied Scientist - Services Special Projects

Join our Apple Services Special Projects Team and help build the next generation of GenAI solutions for millions of people, upholding the user privacy and security that Apple customers know and expect!

Our team shapes the future of Apple's technologies, experiences, and services by building state-of-the-art models, infrastructure, and tools. If you want a collaborative environment where your expertise is valued, and you want to work on hard generative AI and retrieval problems that haven't been solved, this may be a role for you. We are especially interested in scientists who care about what and how models learn, and who use that understanding to make models more accurate, efficient, and robust.

We are seeking a Machine Learning Scientist with a strong track record in applied AI research. You will train and improve the models behind large-scale search, retrieval, and ranking systems, and turn the latest research into production-ready models.

Train and fine-tune retrieval models: dense embedding models, late-interaction (multi-vector) models, and cross-encoder rerankers, with and without instruction conditioning. Fine-tune and adapt LLMs: including small, efficient LLMs for understanding, judgment, and evaluation tasks, using techniques such as LoRA/PEFT, distillation, and preference optimization (RLHF, DPO). Build training data: design training-data pipelines, including contrastive pairs, hard-negative mining, LLM-generated synthetic data, and data-quality analysis. Develop evaluation frameworks: measure retrieval and ranking quality (nDCG, recall, MRR) and model behavior, with automated and reproducible pipelines. Improve efficiency: analyze model internals and use quantization, pruning, and distillation to improve inference speed and cost without losing quality. Scale training and inference: run distributed training and high-throughput inference on GPU clusters, working with engineers on production serving. Collaborate cross-functionally: partner with ML scientists, engineers, and product teams from concept to launch. Communicate: present results to technical and leadership audiences, and contribute to the research community where appropriate.

PhD in Computer Science, Machine Learning, or a related field, or MS with 5+ years of applied ML research experience Strong deep learning and NLP fundamentals, with hands-on experience training and fine-tuning transformer models Proficiency in PyTorch and strong Python programming skills Experience with post-training, fine-tuning, or representation learning for large models Experience designing rigorous experiments and evaluations, and analyzing model behavior Demonstrated ability to turn research ideas into working systems

• Experience training embedding, retrieval, or reranking models (e.g., contrastive learning, late-interaction / ColBERT-style models, cross-encoders)
• Experience with parameter-efficient fine-tuning (LoRA, QLoRA, PEFT) and preference optimization (RLHF, DPO)
• Experience with distributed training (e.g., FSDP, DeepSpeed) and multi-GPU workloads

Experience with inference engines and serving frameworks (vLLM, SGLang, Triton) Experience with model efficiency: quantization, pruning, distillation, or analysis of model internals Experience with data-centric ML, such as data curation, data attribution, or training-data robustness Familiarity with cloud environments (AWS or GCP) and containerization (Docker, Kubernetes) Publications at top venues (e.g., NeurIPS, ICLR, ICML, CVPR, ACL), and service as a reviewer or program committee member