This project aims to create a benchmark and tools for understanding Hindi-English code-mixing, a prevalent yet under-researched phenomenon in Indic social media data. To achieve this, the project will develop a high-quality dataset specifically designed for identifying the language (Hindi or English) of individual words (tokens) in code-mixed text. This dataset will be created by proficient bilingual annotators, ensuring accurate language labels for each word. The project will also compare human annotations against existing automated tools to showcase the benefits of human expertise in providing nuanced and contextually relevant labels with greater accuracy and efficiency. By building this dataset and machine learning models, the project will aid researchers and startups working with large amounts of Hindi-English data. The project has applications in social good areas like hate speech detection and disaster response, and can bridge the gap between language research and policymakers. This project  will not only lead to several publications in top-tier NLP conferences and journals, but will also help create a supportive environment for undergraduate and postgraduate students to develop scientific expertise in understanding the large volume of code-mixed text.