Inspiration

GraphGuard was inspired by the 2024 protests in Bangladesh and the political changes that followed. During that period, there were widespread public discussions and media reports about large amounts of wealth allegedly being transferred out of the country through complex ownership structures and offshore entities. It made me realize how difficult it is to determine who ultimately owns a company or whether hidden ownership chains connect to sanctioned individuals. I wanted to explore how AI and knowledge graphs could make these relationships transparent and easier to investigate.

What it does

GraphGuard is an AI-powered compliance and financial intelligence platform that detects hidden ownership relationships and sanctions exposure using knowledge graphs. Users can explore ownership networks, trace indirect control of companies, identify potential sanctions risks, and receive AI-generated explanations of why an entity has been flagged.

The current prototype uses realistic synthetic data to demonstrate how the platform would operate with real-world datasets.

How we built it

I built GraphGuard as a solo project using a graph-based architecture. I designed an ontology to model people, companies, ownership relationships, and sanctions, then generated synthetic datasets to simulate realistic corporate structures. The backend performs graph queries and logical reasoning to uncover indirect ownership chains, while an AI assistant explains the results in natural language, making complex compliance investigations easier to understand.

Challenges we ran into

The biggest challenge was building every component as a solo developer. Designing an ontology, creating realistic synthetic datasets, implementing graph reasoning, and integrating AI into a single workflow required balancing multiple technologies. Another challenge was demonstrating meaningful ownership investigations without access to confidential or proprietary financial data.

Accomplishments that we're proud of

• Built a complete end-to-end prototype independently.
• Designed a knowledge graph capable of modeling complex ownership structures.
• Created an explainable AI assistant that provides understandable compliance insights instead of black-box results.
• Developed a realistic synthetic dataset that effectively demonstrates ownership and sanctions analysis.
• Combined AI, graph databases, and logical reasoning into a practical compliance solution.

What we learned

This project deepened my understanding of knowledge graphs, ontologies, graph reasoning, and explainable AI. I learned how graph databases can reveal relationships that traditional relational databases often miss, and how AI can make technical compliance investigations more accessible to users. I also gained valuable experience designing and delivering a full-stack project from concept to prototype.

What's next for GraphGuard

The next step is to replace the synthetic dataset with official public datasets from sanctions lists, corporate registries, and beneficial ownership sources. I also plan to expand GraphGuard with real-time sanctions updates, risk scoring, stronger AI reasoning capabilities, interactive graph visualizations, and support for larger enterprise-scale datasets. Ultimately, the goal is to build a platform that helps financial institutions, compliance teams, regulators, and investigative journalists detect hidden ownership networks and reduce financial crime.

Built With

• prometheux
• python