Defog

Defog develops specialized AI models for querying structured data efficiently, offering a fine-tuned large language model called SQLCoder that serves as the industry-leading AI model for text-to-SQL tasks. The platform enables users to get answers from their data without waiting on data teams, supporting drill-down analysis, hypothesis exploration, and follow-up questions through natural language interaction. Defog's technology integrates data across various tools and platforms, querying structured databases from Postgres to Snowflake as well as CSV files. The AI model adapts to user feedback and preferences, providing interpretable and explainable results while ensuring data privacy by never sharing data with anyone, including the AI model itself. SQLCoder has been open-sourced since 2023 with over 300,000 downloads, outperforming major models on text-to-SQL tasks according to industry experts. The company also provides SQLEval, an extensible open-source evaluation framework for measuring model accuracy. Founded in 2023 and headquartered in Singapore, Defog has raised $2.7 million in seed funding from Y Combinator, Script Capital, Hike Ventures, Pioneer Fund, and Davidovs Venture Capital.