Mavenvista Gen AI Report Generation with NLP

Mavenvista GEN AI Case study

Gen AI Report Generation with Natural Language Text

MavenVista Technologies Pvt Ltd (An ISO 27001:2013 and 9001:2015 certified company) is a group of highly passionate and competent procurement & technology professionals committed to delivering intuitive digitization solutions to optimize your direct and indirect spend.

Requirements & Challenges

Mavenvista aims to introduce a Generative AI feature to its customers in the most efficient and cost-effective manner. The goal is to integrate a GEN AI application capable of generating reports from simple text inputs. No filters or selections will be required—just a prompt that generates the needed report.

To illustrate this functionality, here are some sample queries

  • Show me all the titles and details in Gujarat whose rating is more than 9.5.
  • Now show me all the original type titles having ratings more than 7.5 and not in Gujarat.
  • What are the top 10 high value spends?

This feature is expected to be highly demanded as the starting point for GEN AI integration.

The main challenges are that all LLMs are very good in NLP but not that much accurate in formula and calculation.

One of the major concerns is data security and privacy with public LLMs.

Solution to meet challenges

Below is the proposed solution with primary use of AWS Bedrock with Cloude 3 LLM and other services.

A solution designed to satisfy the requirement with Generative BI capabilities on customized data sources (database) hosted on AWS.

    • Text-to-SQL functionality for querying customized data sources using natural language.
    • Performance enhancement through the integration of historical question-answer ranking and entity recognition.
    • Simple agent design interface for handling complex queries through a conversational approach.

Reference Architecture Diagram

The AWS serverless track is selected for the Cloudkida platform

  • AWS Bedrock Titan
  • AWS Bedrock Mistral
  • IAM
  • EC2
  • RDS
  • Dynamodb
  • Opensearch

Custom Module

  • Python based module developed which will convert NLP questions to one or multiple tasks as query to the database.

Third party services used

  • Docker


  • Below are the major benefits of the AWS Generative AI services to the Mavenvista.
    1. Generative BI Capabilities: Leverage advanced Generative BI functionalities tailored to customized data sources hosted on AWS, ensuring insightful and relevant analytics.
    2. Text-to-SQL Functionality: Empower users to query customized databases using natural language, making data access intuitive and user-friendly without the need for SQL knowledge.
    3. Performance Enhancement: Boost efficiency with integrated historical question-answer ranking and entity recognition, ensuring faster and more accurate query responses.
    4. Conversational Query Handling: Simplify complex query handling with a user-friendly agent design interface, enabling seamless interactions through a conversational approach.
    5. Security : Best part of the solution is the complete GEN AI process will be in customer AWS Account only. Any data used in GEN AI process will not be publicly shared at all.


  • Successfully delivered the text-to-SQL with AWS Generative AI services.
  • Additionally it can cache the answer with RAG, increasing efficiency for the repeated queries.
  • Real Time queries unlike many LLM cases where we need to continuously update the knowledge base.

In summery, Electromech achieved, by addressing key challenges to leverage LLM models for SQL queries with required security and privacy.

The integration of Generative BI capabilities on customized data sources hosted on AWS, combined with Text-to-SQL functionality, historical question-answer ranking, and entity recognition, significantly enhances performance and user experience. This solution empowers users to effortlessly query databases using natural language, providing intuitive and efficient data access. Additionally, the simple agent design interface facilitates complex query handling through a conversational approach, making data interactions seamless and user-friendly.