NovaRoute

Inspiration

Customer service times are very long, especially being filled out with questions that can be automated quickly. We wanted to create a model that can analyze user situations and rank customers with priority based on their concerns and situations.

What it does

NovaRoute analyzes a user’s banking activity and fraud risk to determine how their support request should be routed. It connects customers to agents faster when risk is high and lets AI handle simple, low-priority inquiries.

How we built it

We used Streamlit for the interface, FastAPI for the backend, and Gemini AI for message analysis. Fraud detection was modeled using an unsupervised IsolationForest trained on synthetic banking data from Colab.

Challenges we ran into

Setting up ML dependencies locally was a struggle, and integrating real-time fraud updates with Gemini required fine-tuning. Creating realistic, high-risk financial data was also harder than expected.

Accomplishments that we're proud of

We built a working prototype that merges AI reasoning with live fraud analytics in under 24 hours. Watching Gemini dynamically change response priorities as fraud risk increased felt like magic.

What we learned

We learned how to combine structured financial data with generative AI to make intelligent routing decisions. We also saw the importance of explainable results in financial AI systems.

What's next for NovaRoute

We designed the model to be scalable, so in the future, new parameters can be added and passed to the GenAI model to analyze customer sentiment and urgency based on categories beyond just fraud detection.