Background

Are you interested in AI, programming and data analysis and want to apply your skills to real-world traffic safety research?

Our Traffic Safety Research team works with in-depth crash databases to answer complex research questions, with results frequently contributing to scientific publications and conference proceedings. High-quality and accurate data analysis is therefore essential.

Today, R code used in crash-data analyses is manually reviewed by experienced researchers. The review covers areas such as code logic, conditions, filters, data joins, calculations and overall analytical flow. While this provides an important quality check, the process requires considerable manual effort and may not always identify every potential issue.

This Master Thesis explores how Artificial Intelligence and Large Language Models (LLMs) could support and improve this process.

About the Thesis

The objective of the project is to develop and validate an AI-based “Coding Buddy” that can review R code used in crash-data analysis.

Rather than generating new code, the Coding Buddy should act as an intelligent reviewer. Based on the research question, analysis methodology and available domain knowledge, it should identify potential errors or inconsistencies and provide suggestions for improvement.

The Coding Buddy could, for example, review:

Code flow and logic

Data joins and merging

Filters and conditions

Calculations

Missing or incorrect values

Differences between expected and actual analytical outputs

An important part of the project will be enabling the AI assistant to understand domain-specific crash-data knowledge, such as database codebooks, coding standards and relevant literature. Technologies such as Retrieval-Augmented Generation (RAG) or similar AI approaches may therefore be explored.

You will also develop appropriate evaluation metrics and compare the AI-based review with the current manual review process.

Since the project may involve proprietary R code and sensitive crash data, secure data handling will be an important consideration. Where appropriate, anonymized or synthetic data can be used for development and validation.

Research Question

To what extent can an AI-based coding assistant improve the quality and efficiency of R code review in crash-data analysis compared with human review?

Your Project

During the thesis, you will:

Develop an AI-based Coding Buddy for reviewing R code

Define a structured analysis specification and error taxonomy

Build a labelled benchmark dataset of R scripts

Use synthetic data and intentionally introduced coding errors for development and testing

Explore how domain-specific knowledge can be incorporated into the AI solution

Evaluate the Coding Buddy against manual code review

Measure its ability to identify errors and reduce manual review effort

Identify and describe the strengths and limitations of the solution

Deliver a working prototype, thesis report and final presentation

Who Are You?

We are looking for 1–2 Master’s students with an interest in combining software development, data analysis and Artificial Intelligence.

We believe you have:

Experience with R programming

Experience with Python

An interest in AI, machine learning and data analysis

Knowledge or experience in AI/ML, Large Language Models (LLMs) or RAG is considered an advantage.

Most importantly, you are curious, analytical and motivated by the opportunity to turn emerging AI technologies into a practical solution for real-world research.

What You Will Gain

This thesis gives you the opportunity to work at the intersection of Artificial Intelligence, software development and traffic safety research. You will explore how modern AI techniques can be transformed into a practical industrial application while working with a real research use case.

Information and Contact

For more information about the thesis, please contact:

Hanna Jeppsson

hanna.jeppsson@autoliv.com

Junaid Shaikh

junaid.shaikh@autoliv.com

Jawwad Ahmed

jawwad.ahmed@autoliv.com

Jordanka Kovaceva

jordanka.kovaceva@chalmers.se

More lives saved – more life lived!