Master Thesis "Generating Human Readable Software Architecture using Multi-Agent AI"

Problem statement

Context

As software systems become more complex and larger having clear accurate and adaptable architecture diagrams is essential. These diagrams help align development teams communicate design choices and maintain system integrity in distributed environments.

Problem

Automated tools may generate diagrams that are visually unappealing and may lack precision in symbol usage and labeling. Additionally accurately depicting dependencies and relationships between components can be challenging particularly in complex systems with multiple layers. To enhance usability it is essential to translate machine-readable formats into human-readable diagrams ensuring that the resulting visualizations are both clear and easy for users to comprehend.

Proposed solution

This research investigates how multi-agent AI systems can automate and improve the creation of software architecture diagrams. By giving specific roles to different agents such as analyzing requirements modeling components and optimizing visuals these systems can work together to create diagrams that are both technically correct and visually clear. To ensure the information is easy to read its necessary to simplify and group it appropriately while maintaining technical correctness.

Goal:

Develop a prototype using specialized AI agents to analyze software systems and their components to produce clear and easy-to-understand visual diagrams.

Suggested approach:

• The prototype should be run locally using existing multi-agent Bosch tooling that can interact with LLMs as needed.
• Develop the prototype through iterative steps:
  • Select and define personas and their requirements towards content and readability.
  • Rate the generated requirements according to the expectations of the personas.
  • Suggest outcome regarding integration into an agentic workflow.

You will of course have the opportunity to shape the thesis based on your knowledge skills and discoveries made during the project.

Qualifications :

To be successful in the project with think you are:

• A student in Information Technology Computer Science Electronics Math or Physics.
• Required knowledge / courses on data science and AI
• Interested in algorithm development and have some data processing experience with machine learning knowledge.
• Experienced with or have at least some knowledge of programming in Python.
• Self-driven able to challenge yourself and gain the experience needed to move the project forward.
• A person with team spirit social skills and an understanding of software and system architecture.

Additional Information :

Supervisors:  Örjan Percy

We encourage to have a team of 2 master thesis students working on the thesis.

How to apply

Please specify which project you are interested in. Please note: Only applications from students located in the area of Lund are accepted.

Remote Work :

No

Employment Type :

Full-time