A Secure LLM for Enhanced Information Sharing

Azati successfully developed a secure, locally hosted alternative to ChatGPT to improve corporate communication while maintaining strict confidentiality of organizational data.

Discuss an idea

All Technologies Used

PostgreSQL
PostgreSQL
Ruby
Ruby
Redis
Redis
Selenium
Selenium
Sinatra
Sinatra

Motivation

The goal was to automate the data extraction process for a call center that had difficulty manually collecting cellular statistics for hundreds of phone numbers. The customer had a partially functional MVP with significant performance and maintenance issues. Azati’s task was to finalize and optimize the platform to ensure faster data processing, improve accuracy, and minimize ongoing maintenance needs.

Main Challenges

Challenge 1
Frequent Interface Changes

The existing scraping scripts stopped working whenever the telecommunications provider changed the user interface. The solution required ongoing updates to maintain functionality and minimize downtime.

Challenge 2
Performance Issues with Increasing Data

As the number of phone numbers grew, so did the volume of pages to scrape. This caused performance slowdowns, requiring enhanced storage and processing power.

Challenge 3
Fast Data Processing for Daily Reports

The customer needed daily statistics, but the growing number of phone numbers made it difficult to guarantee fast and complete data extraction. Performance optimizations were required to meet deadlines consistently.

Key Features

  • Authentication Module: Bypasses two-factor authentication by automatically intercepting and forwarding SMS codes.
  • Five Scrapers: Each scraper is tailored for a specific telecommunications provider to extract call data.
  • Data Matching and Reporting: A script matches the extracted data with the corresponding phone number and generates daily reports in CSV format.
  • Continuous Monitoring: Daily maintenance checks ensure that all scrapers are functioning correctly and data is being extracted accurately.

Our Approach

Rebuilding and Improving Scrapers
Azati rebuilt and improved the previous scrapers, adding modules for two-factor authentication bypass, handling large-scale data extraction, and ensuring system stability.
Two-Factor Authentication Bypass
A key component was building a background module to intercept and forward SMS codes for automatic two-factor authentication, eliminating manual intervention.
Performance Optimization and Automation
We optimized the system for performance and integrated an automated report generation process to provide the customer with CSV files containing the required phone call data on a daily basis.

Project Impact

Automated Data Collection: Azati’s solution automated the phone call data collection process, drastically reducing manual labor and ensuring that the customer receives daily, accurate reports.

Operational Efficiency: The system has significantly improved the operational efficiency of the call center, allowing for better tracking of employee performance and call quality.

Reduced Manual Labor: Automation of the data collection process freed up employees from routine work, allowing them to focus on more value-added tasks.

Improved Performance: The solution delivers daily, accurate statistics on phone calls, including quality and traffic consumption data.

Scalability: The system is designed to handle an increasing number of phone numbers, with optimizations in place to maintain performance as the call center expands.

Cost Savings: Automation and the reduction of manual maintenance have helped reduce ongoing operational costs for the customer.

Platform Optimization: Rebuilt and optimized the scraping platform, significantly improving stability and performance.

Accurate Reporting: Enabled daily reporting with accurate phone call statistics, including missed calls, call duration, and operator performance.

Time Efficiency: Implemented an automatic two-factor authentication bypass module to save time and ensure smooth operation.

Data Processing Performance: The solution is now able to process data for approximately 96% of the phone numbers required by the customer, with continued performance improvements.

Ready To Get Started

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.