Recent advancements around LLMs and generative AI have laid the groundwork for Agentic AI, where AI agents autonomously carry out tasks by leveraging and communicating with existing tools and systems.
SPARQL is a standard query language used by knowledge graph query engines to retrieve and manipulate data stored as Resource Description Framework triples.
So far, the impact of Agentic AI on SPARQL query engines has not been investigated so far.
The goal of this thesis is to investigate how Agentic AI can help improve the performance of SPARQL query engines.
For example, this can involve an agentic AI loop that continuously analyses SPARQL query logs, and determines if new indexes should be enabled or even disabled.
Furthermore, for query engines such as Comunica that are highly configurable and modular, this also allows AI agents to reconfigure these engines if needed, or even implement custom optimizations on-the-fly, without user intervention.
The student will implement a prototype of an agentic loop on top of the Comunica query engine, and analyze how it can improve continuous performance.