Structuring $30B in Government Procurement: Open Data & AI Architecture for Moroccan Public Tenders

Description and use

An automated pipeline converts thousands of unstructured, bilingual PDF tender archives into structured, searchable data, unlocking a $30 B Moroccan public procurement market for SMEs.

Problem

SMEs and researchers struggled to access and qualify Morocco’s $30B annual public procurement data because it was stored in unstructured, bilingual PDF archives and ZIP bundles with complex regulatory requirements.

Outcome

The pipeline transforms thousands of raw, bilingual tender PDFs into structured, searchable records, enabling SMEs to quickly identify opportunities, comply with new regulations, and analyze market trends. The open dataset and live platform democratize access to Morocco’s $30B procurement market.

Workflow

• Ingest public notices from marchespublics.gov.ma and extract baseline metadata.
• Classify ZIP contents by role (RC, CPS, etc.) to target specific clauses.
• Deterministically extract numeric financial data using regex and LLM validation.
• Cache and serve structured records via Cloudflare edge workers for low‑latency search.
• Publish the resulting dataset to Hugging Face and Zenodo for community use.
• Provide interactive exploration and analytics via Hugging Face Spaces and DuckDB queries.

Evidence

Public procurement in emerging economies represents massive economic volume. In the Kingdom of Morocco alone, government tenders (Marchés Publics) exceed 300 Billion MAD (~$30B USD) every year

Source

To solve this, we built and open-sourced the Morocco Public Procurement Open Benchmark and launched the live intelligence engine at Soumit.ma.

Source