DIC04-NovaMind

Inspiration

Our project, NovaMind, was inspired by the critical "response gap" present in traditional banking systems, where financial institutions are unable to detect and respond to risks in real-time. We recognized the growing importance of ESG (Environmental, Social, and Governance) compliance, but current credit monitoring systems rely on static, historical audits that only capture risks months after they occur. This latency poses significant financial risk. Our goal was to bridge this gap by creating a solution that uses AI-driven, real-time monitoring to help financial institutions stay ahead of potential risks and improve their ESG compliance efforts.

What I Learned

Throughout this project, I gained valuable insights into how artificial intelligence can be integrated with financial compliance. I specifically learned how to leverage real-time data analysis and build an event-driven decision-making system that ensures active risk management. By separating the system into two cognitive layers—the probabilistic perception layer and the deterministic judgment layer—we solved the issue of AI "black box" decisions. This allowed us to ensure transparency and auditability in financial decision-making. Additionally, I learned about modular architecture and how to design scalable, flexible systems that can evolve with regulatory changes.

How the Project Was Built

The project was developed in two primary stages: First, we used Large Language Models (LLMs) to monitor unstructured data from various sources such as news, regulatory filings, and social media. This allowed us to hear and understand risks as they happened. Second, we introduced a deterministic layer to apply pre-defined financial rules for credit score adjustments, ensuring that AI creativity was separated from financial pricing, thus eliminating inaccuracies. We also designed an active response workflow, where the system automatically calculates penalty scores, generates risk memos, and prepares documentation for bank managers to review.

The Challenges Faced

One of the major challenges we faced early on was ensuring that our real-time data processing was both accurate and reliable. Given the complexity of unstructured data, we had to implement algorithms that could detect high-risk events without being overwhelmed by low-impact data. Additionally, AI explainability was a key concern, especially in the financial sector, where decisions must be transparent. We addressed this by decoupling the perception and judgment layers of the AI system, ensuring that every decision made by the AI was aligned with institutional financial policies. Another challenge was integrating ESG compliance into our financial decision-making model, which we tackled by designing real-time feedback mechanisms that enable financial institutions to take proactive measures to mitigate potential risks.

Built With

• amazon-web-services
• github
• javascript
• llms
• mysql
• natural-language-processing
• python
• react
• streamlit
• tensorflow