Postdoctoral Researcher

Lawrence Berkeley National Laboratory • Berkeley, CA

1d • $5,957 - $9,234 • Onsite

About The Position

The Accelerator Physics Group (APG) of the Advanced Light Source (ALS), within the Accelerator Technology & Applied Physics (ATAP) Division at Lawrence Berkeley National Laboratory (LBNL), is hiring a Postdoctoral Researcher to develop agentic artificial intelligence methods and software for accelerator operations. The role combines research on agent architectures, planning, tool use, and evaluation with development of Osprey, the ALS agentic AI platform connecting AI agents with accelerator control systems, facility knowledge, models, and operational tools. The work focuses on enabling agents to perform operational tasks reliably within defined permissions and physical constraints, with appropriate human oversight.

The position is part of MOAT-Core, a multi-laboratory DOE Genesis Mission program focused on AI-driven accelerator facilities. In collaboration with accelerator physicists and partner laboratories, the researcher will develop and validate capabilities for use across the consortium. The role includes participating in ALS-U commissioning activities related to AI systems and contributing to a respectful, collaborative work environment while following LBNL/ALS Environment, Safety, and Health requirements.

This is a full-time, two-year postdoctoral appointment, with possible renewal based on satisfactory performance, funding availability, and operational needs. Work will be performed primarily at Lawrence Berkeley National Lab in Berkeley, California. The salary range is $5,957–$9,234 per month, with an expected starting salary of $8,266 per month or above based on a step schedule. Salary depends on postdoctoral experience after the Ph.D. and is determined by the applicable union contract.

Requirements

• Ph.D. in computer science, physics, engineering, or a related field, awarded within the last three years or expected before the appointment start date.
• Less than three years of paid postdoctoral experience.
• Demonstrated experience designing and building complex software systems in a research environment.
• Proficient Python software engineering skills, including modern development practices such as version control, testing, and documentation.
• Ability to use AI coding tools effectively and critically, including reviewing and validating generated code.
• Working knowledge of machine learning methods and large language models.
• Ability to work independently and effectively in a multidisciplinary team of physicists, software engineers, controls engineers, and operations staff.
• Good written and oral communication skills, including the ability to present research results effectively.
• Ability to foster constructive, respectful, and cooperative workplace relationships.

Nice To Haves

• Research experience with LLM-based agents, tool use, retrieval-augmented generation, or autonomous AI systems.
• Experience at an accelerator facility or with other large-scale scientific instrumentation.
• Experience with agent frameworks and protocols, such as Model Context Protocol, and LLM evaluation.
• Experience with containerization (Docker), CI/CD pipelines, distributed systems, and API design.
• Familiarity with control systems and data acquisition for scientific instrumentation, such as EPICS, Bluesky, and ophyd.
• Contributions to open-source scientific software.

Responsibilities

• Conduct research on agentic AI for scientific facilities, including agent architectures, tool use, planning, multi-agent orchestration, and human-in-the-loop control.
• Design and develop platform components, including knowledge retrieval, tool and skill interfaces to control systems, agent orchestration, and auditing.
• Develop methods for safe and trustworthy autonomous operations, including permission and safety gating, evaluation and benchmarking, and validation against physics and operational constraints.
• Design and develop interfaces connecting accelerator control systems, data archives, logbooks, and digital twins with AI agents.
• Collaborate with accelerator physicists to integrate models and operational workflows as reusable, validated tools for AI agents.
• Collaborate with MOAT-Core partner facilities to integrate and deploy the platform across different accelerators and control systems.
• Use sound software development practices, including testing, CI/CD, documentation, and code review, to produce reliable, maintainable open-source software.
• Disseminate research findings through peer-reviewed publications, conference presentations, workshops, and collaboration meetings.
• Participate in ALS-U commissioning activities related to AI systems.
• Comply with LBNL/ALS Environment, Safety, and Health practices and requirements.
• Contribute to a respectful and collaborative work environment.

Benefits

• Comprehensive medical, dental, and vision coverage.
• Retirement savings through the UC Defined Contribution Safe Harbor Plan, with additional voluntary UC retirement savings options.
• 24 days of Personal Time Off per fiscal year, plus sick leave and paid holidays.
• Eight weeks of Postdoctoral Paid Family Leave for qualifying family and parental needs.
• Berkeley Lab Postdoc Association opportunities for social events, professional development, networking, and scientific exchange.
• Collaborative and inclusive culture supporting research career growth and belonging.

What This Job Offers

• Job Type: Full-time
• Career Level: Entry Level
• Education Level: Ph.D. or professional degree