Duras AI

When banks grant loans, they make decisions based on their complex, proprietary credit rules. These rules are maintained by humans and, given their complexity, even the smallest changes can take up to six months to implement. This lag alone costs banks billions of dollars each year.

At Duras AI, we enable banks to harness the power of AI to update their credit rules and other policy documents. Our platform allows banks to update credit rules faster, more frequently, and on a hyper-local level

What We Learned

Technical Insights:
- [Specific library/framework, e.g., "GOOGLE GEMINI API, REACT"] drastically improved our pipeline efficiency.
- Fine-tuning smaller LLMs (e.g., Mistral) can outperform larger models for niche tasks.
Process Lessons:
- Daily standups and GitHub Projects kept us aligned; wish we’d documented code earlier.
User Needs:
- Early beta testers revealed unexpected pain points (e.g., users wanted [feature]).

*how we built your project*

Step-by-Step JavaScript Implementation:
User-Friendly Interface: Design a sleek, responsive interface for an engaging chat experience, ensuring smooth interactions and real-time responses.
Connecting with the Gemini API 4.Optimizing Performance

Challenges We Ran Into

Model Optimization:
- Balancing accuracy vs. latency (e.g., real-time responses for user queries) was tricky with limited computational resources.
Integration Hurdles:
- Connecting Duras AI’s API with frontend tools (e.g., GEMINI, ) required debugging unexpected compatibility issues.
Team Coordination:
- Divergent ideas on feature prioritization (e.g., NLP refinement vs. expanding use cases) caused short-term delays.

What’s Next for Duras AI

Short-Term:
- Refine MVP based on hackathon feedback (e.g., add [feature]).
Long-Term:
- Expand integrations (e.g., GitHub Copilot-like plugins).
- Explore monetization (e.g., SaaS model for enterprises).

Technical Goals: