VishalNagarkar/banking-sql-intelligence

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

🏦 Banking MIS Modernization: NLP-to-SQL Agent

Automating 60% of PAN India Ad-Hoc Data Requests with High-Speed AI.

🚀 The Problem: Ad-Hoc Backlogs

Large banking organizations receive ~1,000+ ad-hoc data requests daily from nationwide branches. This creates a massive bottleneck for MIS departments, forcing senior developers to spend 60-70% of their time on simple data extraction rather than architectural modernization.

💡 The Solution: AI "First Responder"

This POC demonstrates an AI-driven "First Responder" system that:

• —Offloads 60% of Workload: Handles routine Tier-1 queries (Aggregations, Branch Lists, Balance Checks).
• —Empowers Developers: Leaves complex, high-stakes reporting (40%) to the human experts.
• —Sub-Second Performance: Powered by Groq LPU and Llama 3.3-70B for near-instant results.

🛡️ Security & Architecture

• —Schema Isolation: The AI is provided only with metadata for pre-processed MIS tables, not the live Core Banking System (CBS).
• —PII Masking: Personally Identifiable Information is sanitized at the database view level.
• —Read-Only Interface: The application logic prevents any INSERT, UPDATE, or DELETE operations.
• —Technical Stack: Python, FastAPI, Gradio, Groq API, SQLite.

🛠️ Setup

1. 1.Clone the repo: git clone https://github.com/nagarkarvishal-svg/banking-sql-intelligence/
2. 2.Install requirements: pip install -r requirements.txt
3. 3.Add your GROQ_API_KEY to a .env file.
4. 4.Run: python app.py