Abstract

This project presents an AI-driven system that automates the identification of public contracts that best match a company's profile, optimizing procurement decision-making. By leveraging Google's Gemini AI API (Free Tier), the system analyzes contract descriptions and filters opportunities based on relevance, reducing the time and effort required for manual review. Built with FastAPI, it provides a scalable and modular architecture designed for seamless integration into various workflows. While the current release focuses on contract matching, the platform is structured for future enhancements, including automated contract classification and advanced analytics.

Introduction

This application was originally designed for my mother, who owns a business and is subscribed to a government email list that sends daily notifications about newly opened public contracts at the national level. However, these emails do not only contain newly available contracts; they also include awarded contracts and various public service actions published for transparency purposes.

Due to this, each email can contain more than 800 contract listings, of which only about 100 may be truly relevant as open contracts. Additionally, since the contracts cover the entire country, many of them are outside the geographic scope of her business. Furthermore, a large number of contracts belong to industries completely unrelated to her company's services, making manual filtering even more difficult.

Manually filtering through these large volumes of contracts every day is time-consuming and inefficient. This project introduces an AI-powered contract matching system that automatically processes these emails, extracts the relevant contracts, and filters them based on criteria such as contract status, geographic feasibility, and industry relevance.

Methodology & Architecture

Current Implementation

The contract matching system follows a structured workflow to process government emails and identify the most relevant opportunities:

1. Email Retrieval & Preprocessing:

   • The system connects to an email account and retrieves messages from a specific folder where government contract emails are received.
   • The email content is cleaned and structured, extracting individual contract listings.

2. Filtering & Storage:

   • During preprocessing, contracts that are not open for bidding are filtered out.
   • The remaining contracts are stored in a PostgreSQL database for structured access.

3. AI-Based Contract Matching:

   • The cleaned contracts are sent to Gemini AI, along with predefined instructions regarding the company’s industry and the geographical regions where it operates.
   • Gemini AI analyzes the contracts and ranks them based on their relevance to the business.

4. Frontend & Accessibility:

   • The results are displayed in a React-based frontend, allowing users to efficiently browse and review the most relevant contracts.
   • While this feature is fully functional, the platform is designed for future expansion with additional capabilities.

Technology Stack:

• AI Model: Gemini AI (via Google API) – Used for contract relevance ranking.
• Backend: FastAPI – Ensures high-performance API responses and scalability.
• Database: PostgreSQL with Docker – Stores structured contract data.
• Frontend: React – Fully functional for this feature, with planned expansions.

Results & Benefits

• Significant Time Savings: This system has greatly reduced the time my mother spends searching for relevant contracts. Previously, due to the overwhelming volume of contracts received daily, she often lacked the time to review them thoroughly. Now, she receives a curated selection of the most relevant contracts every day, significantly streamlining the process.
• Improved Accessibility: Instead of manually sifting through hundreds of contracts, she can now focus only on the ones that matter most to her business.
• Potential AI Limitations: While the AI may occasionally miss a contract that could be of interest, the net benefit is still enormous. Given that, in the past, she often skipped reading these emails entirely due to time constraints, the system represents a major improvement in efficiency.

Future Developments

• Frontend Enhancements: While the current frontend is functional, there is room for improvement in terms of user experience and overall usability.
• Prompt Optimization: Further testing and refinement of the AI prompt may be necessary to enhance the accuracy of contract matching and reduce the chances of missing relevant opportunities.
• Additional Features: I am actively working on expanding the application with new functionalities to provide even more value to users.

Repository & Access

The source code is not publicly available yet, but it will be published in the near future. Users interested in updates can follow this project for future releases.

Conclusion

While this may not be the most ambitious project, I consider it to be highly useful. My mother now uses it daily and has told me that it genuinely helps her manage contract opportunities more efficiently. More importantly, this project has achieved something valuable beyond just automation—it has enabled someone who previously struggled to see the short-term benefits of AI to use an AI-powered tool every day and truly understand its usefulness.