Intelligent multi-language AI system for automated ESG data extraction from utility bills

• Existing OCR engine was limited to one document type and a single carbon metric
• No support for multilingual or multi-format document processing
• Each new document type, metric, or language required high-effort manual retraining of models
• Scaling across SME sectors and regions was constrained by costly and resource-intensive processes

‍

• Designed a scalable GenAI-powered system to interpret and extract ESG-relevant data from a wide range of document types and languages, reducing dependency on rigid, template-based models
• Implemented smart filtering and structuring logic to ensure only relevant, high-quality data is captured
• Build and deployed the solution on a secure, cloud-based infrastructure ensuring seamless integration and enterprise-grade performance

‍

• Achieved near-perfect extraction accuracy (99%) significantly reducing manual effort—through the combined power of Azure AI and OpenAI for intelligent document processing
• 50+ language and multi-format support enabled by the flexible model architecture, unlocking adoption across diverse SME sectors and regional markets
• Improved user experience via a drag-and-drop upload interface triggering real-time extraction, powered by the secure and scalable Azure deployment