Master Thesis: Small Language Models for Industrial Operations

About us:

Stegra is on a mission to disrupt the global steel industry by producing green steel. We are building a fully integrated, digitalized and sustainable steel plant in Boden, Sweden, designed to significantly reduce CO₂ emissions compared with traditional steelmaking.

Digitalization, automation and AI are key enablers in how we design and operate our plant. As we move towards operations, we are exploring how emerging AI technologies can support our people, processes and industrial systems in a secure, reliable and scalable way.

This Master Thesis offers an exciting opportunity to explore the potential of Small Language Models (SLMs) in an industrial operations environment and to move from exploration to practical experimentation.

About the assignment

Large Language Models have demonstrated significant potential across a wide range of applications. At the same time, Small Language Models are developing rapidly and may offer advantages for industrial use cases where factors such as speed, cost, privacy, reliability, domain specialization and deployment closer to operational systems are important.

Stegra wants to explore whether, where and how SLM could create value for our operations in Boden.

The purpose of the thesis is to identify and evaluate relevant SLM use cases within industrial operations and investigate the technical and operational feasibility of implementing them at Stegra. For example:

• How do the different real-time requirements of the processes affect applicability?
• How can an operator get help from the "best" colleague to ensure consistent product quality, regardless of who is running the machines on each shift?

The work should combine research and analysis with a hands-on practical implementation. Based on the initial assessment, the student(s) will select one or several promising use cases and develop a prototype or proof of concept to demonstrate and evaluate the potential in a realistic Stegra context.

Potential areas could include, but are not limited to, maintenance, troubleshooting, operator support, safety, technical documentation, knowledge retrieval, classification and extraction of operational information, or AI capabilities deployed close to industrial systems.

An important part of the thesis will also be to understand when an SLM is the right solution - and when another approach, such as a larger language model, traditional machine learning or deterministic software, would be more appropriate.

The work will include:

• Literature review and technology assessment of Small Language Models and their use in industrial environments

• Identification and evaluation of potential use cases within Stegra's operations

• Interviews and workshops with relevant operational and digital stakeholders

• Definition of criteria for evaluating SLM suitability, such as performance, latency, cost, security, privacy, reliability and deployment requirements

• Prioritization and selection of one or several use cases for practical evaluation

• Design and implementation of a prototype/proof of concept

• Evaluation of the prototype in relation to operational needs and alternative technical approaches

• Recommendations for how Stegra could use and scale Small Language Models in future operations

The expected outcome is both a strategic recommendation and a practical demonstration of how Small Language Models could support Stegra's operations.
Who are we looking for

• We are looking for one or two self-driven Master's students studying Computer Science, Software Engineering, Artificial Intelligence, Machine Learning, Data Science, Industrial Engineering or a similar field.
• You should have a strong interest in AI and software development and enjoy combining research with hands-on experimentation. Experience with language models, Python, machine learning, software development or cloud/edge technologies is an advantage.
• Since understanding the real operational environment is an important part of the assignment, you should be able and willing to spend time on site in Boden when needed, working closely with Stegra's Digital and Operations teams.
• We value curiosity, initiative and the ability to independently explore an emerging technology area and turn findings into practical solutions.

Start date: Spring 2027
Duration: 20 weeks (30 ECTS credits)
Location: Boden/Luleå, with presence on site in Boden when required
Number of students: 1-2