Healthcare is increasingly recognized as an important domain for sustainability. Healthcare domain in India and across the globe generates large amounts of unstructured data, including doctors’ notes, research papers, patient files, health related social forum discussions, articles, and tweets. Most of the data lacks meaningful content, limiting its utilization. Standard word models and healthcare dictionaries are insufficient in accurately understanding and extracting medical information from such messy data. A probable solution is to augment the data processing pipeline with knowledge graph (KG), which are a structured representation capturing relationships between entities, concepts, and their attributes. This study proposes a framework using natural language processing and machine learning to build a knowledge graph that extracts useful information from unstructured data. The study uses CORD-19 and MT Sample dataset, which focuses on scientific papers about COVID-19 and medical transcriptions, respectively. The proposed solution maps records to the knowledge graph, thereby linking them to related medical concepts, treatments, diseases, and more. This enables effective querying and the derivation of new insights using appropriate analytical tools. The results confirm that domain-specific pre-training and fine-tuning on biomedical language models like MedBERT are essential for accurate and reliable extraction of clinical entities, making it the recommended choice for real-world biomedical NER applications. Further development and research in this area will explore integrating these extracted entities into more complex knowledge graph structures to facilitate advanced reasoning and decision-making within healthcare domain and sustainable businesses.