Saud Alsulaiman

Leading a project aimed at mitigating Hallucination in large language models. By analyzing and intervening on a model’s activations, I am advancing our understanding of its behavior and enhancing its reliability. This research forms a central component of my master's thesis. Participated in a Kaggle Science Exam competition where I designed a document retrieval algorithm. Utilized SentenceTransformer for embedding textual data and FAISS for efficient context retrieval. This curated knowledge base was then leveraged for multiple-choice question answering using the DeBERTa model.

Supported multiple international gas turbine projects as an Assistant to the Project Engineer, addressing customer specifications and deadlines. Prepared technical documents and data sheets compliant with industry regulations.

Trained AI agents in Unreal Game engine using reinforcement learning to simulate real-world situations such as on-campus traffic and fire drill simulations.