Significant advancements have been made in Natural Language Processing (NLP) in recent years, particularly in Question-Answering (QA). The availability of pre-trained Large Language Models (LLMs) and annotated datasets has driven these improvements. However, most resources are designed for high-resource languages like English, while low-resource languages face challenges due to data limitations. Sinhala, the most widely spoken language in Sri Lanka, with over 20 million speakers, still lacks sufficient annotated datasets and monolingual models for downstream tasks like QA. To address this gap, this study presents a Sinhala QA dataset, SiQuAD, translated from SQuAD v1.1, containing 16,000 unique question-answer pairs. Experiments covering monolingual, cross-lingual, and multilingual approaches are conducted, with the best-performing model achieving an F1 score of 73%, indicating promising capabilities while highlighting room for improvement and future research. The dataset will be made publicly available.