Business Problem

• IVP clients handled terabytes of data covering 1000+ financial metrics. 
• Limited accessibility to raw data for different business users. 
• Lack of a secure and efficient query interface for users with varying levels of data expertise. 
• Interpreting and visualizing the data required high degree of expertise. 
• Executive adoption was low because of the time required to interpret and visualize insights.

About IVP

IVP is a leading financial services firm specializing in investment data analytics, risk management, and regulatory reporting. The company is focused on leveraging cutting-edge technology to enhance data accessibility and decision-making across its teams. 

Solution

GoML enabled IVP to seamlessly integrate their data warehouse with an AI-powered chat interface, enhancing accessibility and decision-making efficiency. By implementing a secure, role-based Q&A system, IVP teams gained structured insights tailored to their needs, improving collaboration and reducing data query errors. 

Data Warehouse Integration: Secure MySQL and Snowflake integration enabling seamless access to raw data.

‍

Web Interface for Collaboration: Streamlined user interaction with chat history and Microsoft Teams integration. (AWS Memory DB for chat storage, Microsoft Teams integration, Azure Bot Services) 

‍

AI-Powered Chat Interface: Users can query data in natural language and receive responses in DESCRIPTIVE formats (actual data table, SQL Code, explanations, visualizations).

‍

Infrastructure: Hosted on AWS for scalability and security. (AWS EC2, Bedrock Agents, Knowledge Base, Flow, Application Load Balancer)

‍

Role-Based Personalization: Tailored insights for leadership teams (descriptive analytics) and data teams (SQL and visualization-based responses). (Python FastAPI for backend logic)

Architecture

• User Interaction Layer 
  React UI application hosted on an EC2 instance for user authentication and query input/output 
  Integration with Microsoft Teams Bot for seamless collaboration (Azure Bot Services) 
• Natural Language Query Processing 
  Converts user queries into structured formats – Dataframes, SQL Query, and Graph Dimensions. 
  Also explains the reason behind reaching a conclusion of choosing that SQL query. 
• Metadata Management & Data Access 
  Metadata stored in DynamoDB for user access control 
  Schema replica maintained in S3 & RDS with metadata enrichment 
  Lambda function and API Gateway trigger schema crawling and updates 
• SQL Query Processing & AI-Powered Insights 
  SQL Retriever Query Engine powered by Claude 3.5 Sonnet (Hosted on AWS BEDROCK) for AI-driven SQL generation 
  Persona-based filtering ensures tailored responses for different users 
  Business context-aware query processing for accurate insights 
• User Access & Security 
  Access Filter and User Query Clean module on EC2 for secure role-based access 
  Logs maintained for tracking user interactions and data requests 
• Collaboration & Chat History 
  Chat history stored in AWS Memory DB for future and QUICK reference 
  Microsoft Bot Framework enables adaptive card responses for visualization, tables, SQL (Azure Bot Services) and explanations. 
• Client SQL Server Integration 
  Secure connection to the Client SQL Server for structured data retrieval 
  Ensures seamless data access while maintaining compliance