Staff Scientist - Post-Training and Reinforcement Learning for AI for Science

About the role

The Argonne Leadership Computing Facility (ALCF) is seeking a Staff Scientist to advance the next generation of foundation models and learning systems for scientific discovery. This role focuses on research and development of post-training methods, including reinforcement learning, preference optimization, adaptation, and alignment techniques, for scientific AI models and workflows.

Responsibilities

• Conduct research and development aligned with Argonne's strategic mission in computation, AI, and scientific discovery.
• Develop, scale, and optimize post-training methods for scientific foundation models.
• Advance techniques that improve the performance, controllability, reliability, and scientific utility of AI models for science applications.
• Design and evaluate methods for applying reinforcement learning and post-training pipelines to large-scale scientific and data-intensive environments.
• Develop and optimize workflows for training and post-training on leadership-class supercomputers and emerging AI-oriented architectures.
• Partner with computational scientists, applied mathematicians, and domain researchers to apply foundation models and adaptive learning systems to challenging scientific problems.
• Address algorithmic, systems, and data challenges associated with large-scale training and post-training.
• Conduct original research in computational science and AI at scale, and communicate findings through publications and other research outputs.
• Collaborate with colleagues across national laboratories, universities, industry, and supercomputing centers.

Requirements

• Bachelor's degree and 5+ years of experience, or a Master's and 3+ years of experience, or a PhD, or equivalent.
• Education in computer science, applied mathematics, statistics, computational science, or a related field.
• Demonstrated advanced knowledge in machine learning, reinforcement learning, large-scale model training, post-training, optimization, data mining, or statistics.
• Advanced knowledge and significant programming experience in Python, C, or C++.
• Significant experience with machine learning frameworks such as PyTorch or JAX.
• Experience with large-scale training, distributed learning systems, or post-training workflows.
• Experience with software development practices for computational science and machine learning systems.
• Able to work effectively in interdisciplinary teams.
• Effective written and verbal communication skills.
• Able to model Argonne's core values of impact, safety, respect, integrity, and teamwork.

Qualifications

• Experience with reinforcement learning, policy optimization, bandits, preference learning, or related methods.
• Experience with post-training methods for large models, including supervised fine-tuning, reinforcement learning from feedback, direct preference optimization, reward modeling, or model adaptation.
• Experience with distributed training, large-scale optimization, and multi-node or multi-accelerator execution.