Sourish-Kanna/CenQuery

CenQuery is a niche application of existing NL-to-SQL fine-tuning techniques (QLoRA on Llama-3). While focusing on the Indian Census provides a specific use case, the low star count, lack of velocity, and reliance on standard fine-tuning patterns make it easily reproducible. Furthermore, frontier models (GPT-4o, Claude 3.5) demonstrate high zero-shot performance on SQL tasks when provided with schema context, significantly reducing the need for domain-specific fine-tuned 8B models.