Master Thesis AgentOPS frameworks and infrastructure

Problem statement

Generative AI (GenAI) is transforming software development. Tool support powered by large language models (LLMs) such as GitHub co-pilot and ChatGPT are becoming an integrated part of the software development workflow. LLM-based agents which take on software development tasks more autonomously are now starting to emerge. While an LLM provides replies to prompts in one step an LLM-based agent can autonomously perform tasks and take decisions using an LLM in several addition to an LLM the agent includes perception capabilities memory and supporting resources and can provide the LLM with context intelligence via the Model Context Protocol (MCP).

When multiple agents collaborate we get a dynamic system where agents jointly work on a problem where human involvement changes from being aware of all details to a supervisory role. Setting up an infrastructure for a multi agent system and how to manage it is still an open question where several frameworks and platforms are emerging and key characteristics and how to balance different decisions are not known.

Proposed solution

We want to explore which are the prominent frameworks and platforms emerging for managing the agent operations to understand the capabilities and limitations. We also want to set up an example multi-agent system on a Bosch infrastructure to get hands-on experience and an insight into what is required in terms of knowledge and infrastructure to operate a multi-agent system.

We also want to explore and understand where to deploy agents on a personal computer on a local on-premises hardware or in the cloud. There are different aspects to consider such as cost and data integrity. There might be decision to take while setting up a system (design-term) as well as more or less automatic decisions during runtime (operations).

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

Scope of thesis project

Two students on master level working onsite at the Lund office (primarily).

Qualifications :

Your profile

In order to be successful in the project we think you are:

• A student in Information Technology Computer Science Electrical Engineering or similar
• Experience with AI a general and specifically using pre-trained models and related technique (e.g. RAG multi-agents etc.)
• Experience and interest in agentic development
• Driven able to challenge yourself and gain the experience needed to move the project forward.

Additional Information :

Supervisor

• Thomas Olsson

Please note: Only applications from students at a Swedish University are accepted.

Remote Work :

No

Employment Type :

Part-time