Master Thesis "Enhancing a Log-Analyzer with a Knowledge Graph"

Company Description

Welcome to a world, where your ideas lead to something big. Welcome to Bosch.
At Bosch, we shape the future by inventing high-quality technologies and services that spark enthusiasm and enrich people’s lives. Our promise to our associates is rock-solid: we grow together, we enjoy our work, and we inspire each other. Join in and feel the difference.

Bosch R&D Center Lund stands for modern development in cutting edge technology in the areas of connectivity, security, mobility solutions and AI. We are growing rapidly and looking for people to join us on our mission to become the Bosch Group’s 1st address for secure connected mobility solutions. We are working on a range of interesting projects, with a particular focus on software development for the automotive industry, electrical bicycles and Internet of Things.

Job Description

Problem statement

Context

Large Language Models (LLM) can serve many useful purposes for large enterprises by enabling easy and approachable access to large sets of data and provide higher efficiency in product development. Bosch already has AI based tooling to support our software developers and we explore new methods to optimize our supporting tools.

Problem

The results from the LLM, however, can be inaccurate, meaning it can sometimes provide incorrect answers to questions or is unable to correctly interpret the question or task. The LLM can also produce results that are made up of non-existing data, so-called hallucinations. These issues can have severe impact on the application/developers that rely on them, making them unreliable and invalid for production level solutions.

Proposed solution

One approach to improve accuracy and reduce hallucinations is to enhance the LLM by storing the underlying data in a knowledge graph. The knowledge graph provides semantics and context to the data, clearly labeling the data points and the relations between them. This makes it easier for the LLM to draw the correct conclusions from the data, because of how the knowledge graph structures the data. The solution would be implemented as a proof of concept and evaluated based on the performance regarding accuracy and hallucinations. The solution shall be tested in an existing LLM based Bosch Log-Analyzer tooling.

Goal of the Master Thesis:

To implement a proof-of-concept for a Graph RAG LLM, identify how to evaluate the accuracy and hallucinations and evaluate the performance of various queries.

Suggested approach:

• Implement the GraphRAG with use of an existing LLM
• The LLM should be extended with a GraphRAG implementation
• As data source for development GraphRAG an open-source dataset should be used
• Validation shall be done with Bosch Log-Analyzer tooling

For example; use of NASA Open Data Portal (various open data sources provided by NASA) logpai/loghub (a large collection of system log datasets for AI-driven log analytics)

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

Qualifications

Your profile

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

• A student in Information Technology, Computer Science, Math, or Physics.
• Required knowledge: courses on data science, AI, and graph databases.
• 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, ability to challenge yourself, and gain the experience needed to move the project forward.
• A person with team spirit, social skills, and curiosity for exploring new technology areas.

Additional Information

Scope of master thesis project

1-2 students completing 30 credits each (20 weeks) onsite at the Lund office.

How to apply

Please specify which project you are interested in. Please note: Only applications from students at a Swedish University are accepted.