Bringing Clarity to Government Procurement with AI

Building a Government Transparency Platform for Public Accountability

We built MARIA to simplify how procurement data from the Dominican Republic is collected, analyzed, and shared. Before MARIA, users had to visit government websites, download documents, and read long files manually to find bidder names, contract amounts, award details, and company information.

MARIA brings all of this into one searchable platform, helping users track tenders, companies, ownership links, and spending trends faster.

The Challenge

MARIA had to solve the main problems that made procurement data hard to use:

• Tender data was scattered across different public sources
• Documents came in mixed formats like PDFs, web pages, and scanned files
• Manual tracking was slow and often missed important updates
• Company names appeared in different spellings or formats
• Ownership links between people and companies were difficult to follow
• Bid amounts, award details, and agency names were often buried in long documents
• Large document volumes made manual review impossible
• Duplicate records could affect data quality
• The platform needed to update often without losing information

The goal was to turn scattered public records into clean, searchable, and useful procurement intelligence.

Our Solutions

We built an automated system that collects, analyzes, and organizes procurement data every day. MARIA gathers tender documents from public sources, stores them securely, and uses AI automation to extract key details such as bidders, award values, public bodies, dates, and process types. The platform also cleans and matches company records to reduce duplicates. Users can then search tenders, view company profiles, explore ownership links, and generate reports through a simple dashboard.

Results

• Daily tender collection became automated
• Manual monitoring was replaced with near-complete procurement coverage
• New procurement data became available within 24 hours
• 12,000+ tenders were indexed and made searchable
• 50,000+ companies were organized into company profiles
• 100,000+ bidder records were structured for analysis
• AI extraction reached 92% accuracy in sample checks
• Job processing achieved a 99.2% success rate
• Searches returned results in under 200ms
• Reports and exports became available instantly