Thesis: AI-Based Information Retrieval – A Comparison of Traditional and Modern Methods

Advances in generative AI have expanded the options for searching and interpreting text, but traditional search can still be the better choice in some situations. This thesis compares keyword search, embedding-based semantic search, and retrieval-augmented generation (RAG) on unstructured text. The aim is to establish when AI-based retrieval delivers worthwhile improvements in quality, speed, and cost.

The assignment

• Build three retrieval solutions using public text datasets: keyword-based search, semantic search with vector embeddings, and RAG using a large language model.
• Evaluate the solutions against the same datasets and questions, measuring retrieval relevance, answer correctness, response times, and computational costs.
• Investigate how often each approach returns incorrect or misleading information and which types of questions benefit from an LLM-based solution.
• Potentially explore how the amount of retrieved context influences answer quality and the risk of misleading outputs.
• Summarize the strengths, limitations, and practical trade-offs of the methods to support organizations choosing a search approach.

Who we are looking for

• Bachelor’s or master’s student in computer science or computer engineering, expecting to graduate in June 2027.
• Fluent in both Swedish and English, in speech and writing.

Practical details

Based at Knightec Group’s Sundsvall office, starting in January or March 2027. Apply through the company’s careers page by November 30, 2026; early applications are encouraged. Questions can be sent to johanna.edstrom@knightec.se.