Imagine trying to ask complex questions about data in a language that most AI models don't fully understand. That’s the challenge researchers tackled for Indic languages like Bengali and Hindi with the introduction of TableQA. TableQA systems aim to answer questions using information stored in tables, going beyond simple lookups to perform calculations and comparisons. Researchers developed a clever system to automatically generate large datasets for training these models, overcoming the lack of readily available labeled information. The system starts by extracting tables from Wikipedia in the target language, creating a rich source of culturally relevant data. Then, it generates both SQL-like queries and corresponding natural language questions. By executing these queries, the system extracts answers, creating a powerful training loop. The results are impressive: models trained on this data beat state-of-the-art large language models (LLMs), proving the effectiveness of this approach. The models even displayed the ability to perform mathematical operations, demonstrating a deeper level of table comprehension. Interestingly, a model trained on Bengali performed surprisingly well on Hindi, even though the scripts are different, showcasing the potential for cross-lingual application. This breakthrough opens exciting doors for enhanced information access, data analysis, and more in these widely spoken languages.