Agent-based performance assessment - Task 046

Fit to programme

This task has been identified by the working groups as part of the agenda behind WP 1.2.

The task number is 046.

Summary

Running a performance analysis along the lines of SHAREing’s high-level assessment is a rather mechanical task: for each of the five high-level dimensions, certain measurements must be taken, and the resulting data then feeds into a report, supplemented with additional contextual information.

Many of these steps can likely be automated. We suggest that a project addressing this task investigate how to use agentic AI to automate most of the assessment steps: what are the typical prompts that would delegate most of this work — for example, running scalability studies — to an agent, in order to streamline the evaluation?

Methodology

Use agentic AI to study the performance of one code and conduct a high-level assessment along the lines of the SHAREing high-level methodology. Document the individual prompt entries and summarise them in a report/blog.

Outcomes

Blog/report with instructions how to make an AI do lots of the technical work of a high-level assessment.