Location: Upton, NY
Cap-Exempt H-1B Position — No lottery required
**Position Description** The REDWOOD project funded by DOE ASCR is meant to study the challenges of scientific payloads resilience in the heterogeneous distributed computing in environments such as those in HEP, NP, astro-particle physics, and fusion science, as well as other data intensive physics and astrophysics projects. The program of work proposed herein is carried out under the REDWOOD project. It consists of three main thrusts: The first thrust is maintenance, development, support, and exploitation of a simulation tool known as CGSim, whose goal is the simulation of distributed computing systems. The task associated with this is the overall responsibility for the CGSim project, including deployment and documentation. Under this program of work, CGSim would be expanded to incorporate more grid sites as the descriptive data for more sites becomes available. The grid simulation would be used to test new, AI-based job scheduling algorithms. The second thrust is to aid in the demonstration of large language systems AI tools to monitor distributed computer systems. A current prototype system is referred to as AskPanDA. It aims to provide monitoring feedback to users of all experience levels in the form of natural language answers to natural language questions, generating all the necessary database queries, using Model Context Protocol (MCP) technology to access large language models (LLMs). Under this program of work the monitoring system should be demonstrated on the CGSim-simulation of distributed workflow management systems, allowing for further development of AskPanDA-like systems in a nondisruptive, controlled environment. A third thrust is the support of synergistic activities at Brookhaven aimed at AI-based simulation of the workflow management system, and studies of AI-based tools for resilient workflow management. This is under the rubric of the simulation and modeling (ModSim) track of the REDWOOD project. The impact on novel workflow scheduling algorith