AI/ML Researcher/Engineer - Foundation Model Post-Training

Summary

We are a tight-knit group of researchers and engineers responsible for building large scale frontier foundation models at Apple. We believe the most interesting breakthroughs in deep learning happen when we bridge the gap between raw model capability and user-centric utility.

Minimum Qualifications

Demonstrated expertise in deep learning with a focus on LLMs, post-training, or reinforcement learning, backed by a strong record of academic or real-world accomplishments in these or closely related domains. Proficient programming skills in Python and a major deep learning framework such as JAX or PyTorch. Masters/PhD, or equivalent practical experience, in Computer Science, Machine Learning, or a related technical field.

Preferred Qualifications

Experience training state-of-the-art large models at scale, with familiarity in distributed training challenges and trade-offs. Experience improving model performance on complex reasoning tasks (math, coding, logic). Experience with various transformers architectures and its transformations. Strong communication skills and a passion for working cross-functionally across Research and Product teams.

Description

In this role, you will play a critical role shaping the future of our LLM efforts, specifically in transforming our models into highly capable, intelligent assistants that power billions of Apple products. You will tackle core training challenges in instruction following, tool use, deep reasoning, and architectural adaption — designing models that deliver magical, deeply integrated, and privacy-forward experiences across the Apple ecosystem. You will work alongside a fast-growing team of world-class experts to explore novel training strategies, architectural adaptations, and advanced evaluation methodologies.

Responsibilities

Design and iterate on end-to-end post-training strategies (including Reinforcement Learning) to unlock model capacities toward achieving specific model behaviors. Pioneer novel algorithms for preference optimization, model steering, and safety. Drive our data strategy by researching methods for high-quality human and synthetic data generation, automated data filtering, and curriculum learning to improve instruction following and reasoning. Design robust evaluation methodologies to measure model helpfulness, factuality, and utility, moving beyond static benchmarks to accurately capture real-world performance. Partner closely with pre-training teams to inform architecture choices, and with product teams to translate user requirements into model capabilities.