nasa-smd-ibm-st is a sentence embedding model based on Bi-encoder, fine-tuned from the nasa-smd-ibm-v0.1 encoder model. It was trained on 271 million training samples, including 2.6 million domain-specific samples from NASA Science Mission Directorate (SMD) documents. This model aims to enhance natural language technologies, such as information retrieval and intelligent search, for applications in SMD's natural language processing tasks. It can be broadly used in information retrieval, sentence similarity search, and other NASA SMD-related scientific use cases.