GitHub is advancing the field of semantic code search by leveraging deep learning to improve the discoverability of code beyond traditional keyword searches. By developing machine learning models that represent code and text within a shared vector space, GitHub aims to facilitate more intuitive searches where even non-matching keywords can yield relevant results. The approach involves training sequence-to-sequence models to create code summaries and fine-tuning these models to align code representations with textual descriptions. This semantic search capability could significantly enhance onboarding processes for new developers and improve code accessibility. Although the current implementation demonstrates the potential of semantic search within a limited dataset, ongoing research is focused on expanding its scope and refining its components, such as data preparation, model architecture, and evaluation methods. GitHub's efforts are part of a broader initiative to explore new use cases, including multilingual code searches, and they are inviting collaboration from interested researchers and developers.