DMSPN

Audit tool for banking institutions to detect fraud and money laundering, scoped to small and medium (PYMEs) restaurants in Mexico.

Comment

Inspiration

Fraud and money laundering represent critical threats to small and medium enterprises (SMEs). Unlike enterprise-level financial institutions, small businesses often lack the specialized compliance infrastructure and dedicated risk teams needed to spot sophisticated financial crime. This leaves them exceptionally vulnerable, as fraudulent patterns frequently blend into daily operations unnoticed until severe financial damage has occurred.

What it does

DMSPN is an intelligent risk-analysis platform designed to streamline financial oversight and accelerate fraud detection. By combining automated data aggregation with AI-driven insights, the platform helps risk analysts identify suspicious transaction patterns, flag potential money laundering schemes, and prioritize threat levels in real time. This drastically reduces manual review times and empowers teams to mitigate risk before it escalates.

How we built it

Our solution relies on a modern, cloud-native architecture optimized for speed and analytical power:

  • Data & Storage: Built on Snowflake to handle complex financial queries and process high-volume transactional data with high performance.
  • Frontend: Developed as a responsive web application using Next.js and React, providing an intuitive, dashboard-first user experience.
  • AI Orchestration: Integrated Google Gemini as an intelligent orchestrator driving a specialized Snowflake Agent. This allows analysts to query complex relational data using natural language and receive contextualized threat assessments instantly.

Challenges we ran into

Integrating cutting-edge AI frameworks with cloud-native data platforms presented a steep learning curve. We had to quickly adapt to emerging libraries while architecting a resilient system design. Ensuring low-latency communication between the Gemini agent and Snowflake required careful optimization to avoid query timeouts and keep the UI responsive during live data fetch operations.

Accomplishments that we're proud of

We successfully designed, built, and deployed a fully functional end-to-end prototype within a tight timeframe. We are particularly proud of seamlessly connecting our Next.js frontend to an AI agent that dynamically queries Snowflake backend tables, effectively bringing our initial vision of automated risk analysis to life.

What we learned

This project deepened our technical skills across several domains. We gained hands-on experience in modern web-based data visualization, learning how to present complex financial patterns through intuitive dashboards. Additionally, we mastered agentic workflows, prompt routing, and optimal state management between serverless APIs and dynamic frontend components.

What's next for DMSPN

  • Scalability: Upgrade the data pipeline to handle higher transaction throughput without performance degradation.
  • Real-Time Streaming: Transition from periodic database polling to a real-time event-driven architecture (e.g., streaming data pipelines) for instant alerting on anomalous transactions.
  • Advanced ML Models: Introduce custom anomaly-detection models directly into Snowflake to complement Gemini's NLP capabilities.

Mesa 165 built alongside Bruno Arroyo del Castillo and Mateo Zamora Grajeda

Built With

Share this project:

Updates