Inspiration
With growing uncertainty in global markets, we realized that most retail investors have no clear way to evaluate how fragile their bond portfolios are. Traditional tools show past data, but not how risks can spread or escalate. We wanted to empower users to understand systemic vulnerabilities—visually and intelligently—so they can invest with more confidence.
What it does
Our tool analyzes a user’s bond portfolio and highlights hidden risks. It simulates how economic shocks might trigger defaults across companies, sectors, and regions. Users can explore their portfolio through an interactive bubble dashboard, see which areas are most exposed, and get AI-powered recommendations based on potential cascading effects.
How we built it
We modeled individual default risks using Cox's proportional hazards model and generated joint default scenarios using copula functions. Then, we built a synthetic dataset to simulate realistic stress conditions. A Graph Neural Network (GNN) was trained to detect patterns of systemic risk. Finally, we built an interactive React dashboard that presents this information in a user-friendly way, suitable for integration into Revolut's "Invest" section.
Challenges we ran into
One major challenge was creating a realistic synthetic dataset that accurately reflects market dynamics. Designing the GNN to pick up on meaningful default patterns across company networks also took trial and error. On the frontend side, it was tricky to balance visual clarity with complexity—especially when displaying risk breakdowns by multiple categories.
Accomplishments that we're proud of
We're proud of turning advanced financial theory and machine learning into something accessible for everyday investors. From modeling simultaneous defaults to building a responsive and intuitive interface, we created a full pipeline from data to insights to user interaction.
What we learned
We deepened our understanding of systemic risk modeling, survival theory, and copula functions. We also gained hands-on experience training GNNs on structured financial data. Most importantly, we learned how to simplify complex analytics into visuals and insights that anyone can use to make better investment decisions.
What's next for Bond-based portfolio risk reviewer
Next, we want to connect the tool to real market data and test it with live portfolios. We'd also like to simulate how real-time news events (e.g., a rating downgrade) could dynamically update risk estimates. Finally, we aim to integrate this tool fully into a mobile platform like Revolut, bringing systemic risk awareness directly to retail investors.
Log in or sign up for Devpost to join the conversation.