preloader
AI_Agent_Personal_Data_Audit_Overview

AI Agent - Personal Data Audit

This is an AI-powered content moderation tool for reviewing articles and submissions, detecting personal information, violent language, or discriminatory terms, and highlighting them quickly for editors.

Project Overview

Client

Client name: Private
Nation: Private

Industry

Private

Platform

Web Application

Team Size

Private

Technologies

Private

Features & Security

AI_Agent_Personal_Data_Audit_Key_Features

Functional

  • Efficient upload and processing of news content, including PDF documents.
  • AI-powered analysis to identify sentiment, key individuals, and potential reputational risks.
  • Automated tagging and classification to surface relevant insights.
  • Moderation interface for reviewing and refining AI-generated results.
  • Secure role-based access for analysts and moderators.
  • Comprehensive audit trails for all changes and user actions.
  • Seamless integration of AI insights into the content review workflow.
  • Scalable and modular system designed to support future enhancements and AI upgrades.
  • User authentication via AWS Cognito.

Non-Functional

  • High Performance Auto-scaling server system ensures smooth operation with tens of thousands of concurrent users
  • Web Responsive Ensuring optimal system performance across different web platforms and devices
  • User Experience Clear warnings and prevention of transactions if the minimum amount received is significantly lower than expected, avoiding user confusion and financial losses

System Security

  • HMAC Authentication Verifies that requests are coming from expected sources and have not been tampered with during transit
  • DDoS Attack Prevention  Utilizing CloudFlare, the system can monitor and quickly block DDoS attacks based on request IP addresses
  • XSS, CSRF & SSRF protection
  • Encrypted data User’s sensitive information related to banking, phone numbers, addresses, etc., is encrypted in the database
  • CloudFlare (CDN) integration  Integration with CloudFlare enables easy prevention of DDOS attacks, unauthorized access through OTP verification, and real-time security monitoring
  • IP Whitelist/VPN Access to the web admin system is restricted to IP Whitelist or authorized VPN accounts
  • Secret key stored in Vault All project keys (such as secret keys, third-party API keys) are stored securely in a separate storage system (without any information stored in the database or source code)
  • Multisignature cold wallets
  • Hide sensitive data in logs All important information is hidden in the system logs
  • Frequent security scanning with OWASP, SynkIO, and AWS Security Scans Integration of well-known security scanning tools such as OWASP, AWS Security Scans, and mandatory SEC tools ensures the detection of security issues and daily updates on new security vulnerabilities

User Security

  • Two-factor Authentication Integration of One-Time Password (Google Authenticator) for functions such as FIAT/Crypto Withdrawal and User Security Features
  • Login Throttling Limits the number of logins attempts an attacker can make while providing multiple opportunities for users to remember their passwords
  • Complexity policy for password and PIN Ensures a certain level of complexity for passwords and PINs (for mobile phones) to prevent easy guessing by hackers

Solutions

Decoupled upload system with asynchronous processing: Implement a queue-based upload system using AWS S3 and Lambda to store files and trigger processing asynchronously.

Batch processing and page limitation: Handle documents in batches (e.g., page-by-page processing) or restrict the number of pages per request to maintain performance.

Active learning with feedback loop: Apply an active or continual learning pipeline, storing user annotations in a feedback database and retraining the model in batches with version control.

dexwonder-project-solution

Challenges

The project team encountered no technical challenges, including achieving the system’s functional criteria as specified by the SEC, based on prior experience working on other exchange systems. However, the project has to face challenges due to the Cryptocurrency downtrend:

Financial Impact of Cryptocurrency Downtrend

Risk of Uncertain Market Conditions

Balancing Project Quality and operational Cost

Managing Stakeholder Expectations

Ensure legal protection for users set by the Securities and Exchange Commission (SEC)

Cost & Quality Optimization

The way we save cost and optimize quality for this project:

Leverage our experience from previous AI and data-related projects to accelerate development while maintaining high quality.

Optimize AI Agent workflows and data processing to reduce resource consumption and improve overall system performance.

Utilizing AI in the workflow

With the current capabilities of AI in assisting various tasks, applying AI to tasks such as UI design, testcase review, and source code review has helped the project team save a significant amount of time and effort.

Automation of code review

Using SonarQube for automated code review helps save costs associated with manual code review and ensures the quality of the source code. 

Proper task allocation

Each project involves both experienced and new team members. Although all team members receive training before joining the project, it is not possible to cover the entire system comprehensively. Therefore, to accelerate project progress and ensure the highest quality, it is essential to assign experienced individuals to tasks that require expertise in a specific domain. For example, when developing a Mobile App, experienced individuals can handle API integration for screens with complex logic, while newcomers can focus on UI development, requiring little or no experience in Crypto Exchange.

Adhering to strict quality processes

It may seem difficult to persuade, but usingthird-party services can save more money than handling a service internally. For instance, by using a third-party service for Slider Captcha (a service provided by a dedicated team), the project can save costs compared to implementing and maintaining a separate service. Based on the company's experience and project team's expertise, they have identified third-party providers with the best cost-effectiveness, stable service quality, and sustainability.

Utilizing third-party services to save costs

Each error results in a series of related actions, such as logging bugs, fixing and retesting, reporting, and capturing evidence. By ensuring strict quality processes within the company, the project team has minimized the occurrence of errors, reduced rework time and saving costs.

Experience from previous projects

With their own experience and a team of experienced consultants, the project's technical and business implementations are streamlined and accurate, minimizing costs, time, and resources. For a large-scale project, even a single incorrect business or technical approach can lead to a cascade of issues. By forecasting and providing appropriate solutions based on previous project experiences, this project ensures the highest quality while maintaining reasonable and cost-effective measures.

Accomplishment

The project exceeded expectations and was successfully delivered within 2 months. In addition to developing the AI Agent for Personal Data Audit as required, we also provided valuable recommendations and support throughout the project, including optimizing the audit process, improving data analysis accuracy, and suggesting practical solutions to enhance the overall system’s performance and usability.

Need a companion?

Don’t Hesitate, Contact Us Now!

    Your Chat Channel
    (Select the chat platform you are using)

    Let's work together!