Research Aide - CPS - Li, Jinchu - 3.19.26.

Argonne National Laboratory•Lemont, IL

2d•$27 - $40

About The Position

A major challenge across computational science is that much of the existing HPC simulation ecosystem remains effectively inaccessible to agentic AI systems. Legacy codes are typically exposed through complex command-line interfaces, rigid input specifications, and manually configured software environments, creating a high barrier for integration with LLM-driven workflows. While the Model Context Protocol (MCP) provides a standardized interface for connecting language models to scientific tools, the current practice of manually implementing an MCP server for each simulation package. This project addresses this limitation by developing MCP-Forge-for-Science, an automated, generative framework in which an LLM analyzes software documentation, source code, and usage examples to synthesize standardized MCP servers directly. By automating tool wrapping and interface generation, MCP-Forge enables rapid, scalable onboarding of existing HPC software into agentic AI workflows. We have developed a semi-automated pipeline for constructing MCP servers for a limited set of simulation libraries. The primary focus of this project is to extend this effort into a fully automated MCP synthesis framework, MCP-Forge-for-Science. The student will design and implement a robust validation suite that executes test simulations to verify correct flag mapping, command invocation, and input handling, ensuring that automatically generated MCP servers are reliable and free from hallucinations prior to production deployment. In parallel, the student will systematically evaluate MCP servers generated by both open-weight models hosted on ALCF inference endpoints and proprietary models accessed via Argo, comparing their accuracy, robustness, and synthesis efficiency. The validated MCP servers will be integrated with HPC-native workflow managers such as Parsl and Ensemble Launcher, enabling agentic frameworks to orchestrate legacy simulation software at scale on ALCF’s current and future systems.

Requirements

The entirety of the appointment must be conducted within the United States.
Applicants must be:
Currently enrolled in undergraduate or graduate studies at an accredited institution.
Graduated from an accredited institution within the past 3 months; or
Actively enrolled in a graduate program at an accredited institution.
Must be 18 years or older at the time the appointment begins.
Must possess a cumulative GPA of 3.0 on a 4.0 scale.
If accepting an offer, candidates may be required to complete pre-employment drug testing based on appointment length. All students remain subject to applicable drug testing policies.
Must complete a satisfactory background check.

Responsibilities

The student will design and implement a robust validation suite that executes test simulations to verify correct flag mapping, command invocation, and input handling, ensuring that automatically generated MCP servers are reliable and free from hallucinations prior to production deployment.
In parallel, the student will systematically evaluate MCP servers generated by both open-weight models hosted on ALCF inference endpoints and proprietary models accessed via Argo, comparing their accuracy, robustness, and synthesis efficiency.
The validated MCP servers will be integrated with HPC-native workflow managers such as Parsl and Ensemble Launcher, enabling agentic frameworks to orchestrate legacy simulation software at scale on ALCF’s current and future systems.