Services
Generative AI Case Study
Financial B2B Credit Risk Deals Analysis
Business Challenge
- Manual analysis of unstructured documents for extracting risk-related insights was time-consuming and required a dedicated team.
- Hundreds of deal memos were reviewed manually, leading to inefficiencies.
- Lack of historical records and control over documentation and analyzed insights.
Solution
- Infrastructure: Fully hosted on-prem, with the exception of Claude Sonnet 3.5.
- MinIO for Object Storage (AWS S3 compatible).
- Embedding Model served via LangServe in Kubernetes.
- Qdrant Vector Store deployed in Kubernetes.
- Entire application hosted within Kubernetes.
- Functionality:
- Insights processed into a database with a custom UI.
- LLM self-evaluation and correction, enhanced with human feedback.Functionality:
Key Outcomes
- Efficiency: Reduced document review time by up to 90%.
- Control: Business SMEs manage the solution, providing full oversight of AI-generated results.
- Scalability: The solution is reusable, enabling quick onboarding of additional departments.