Client:
One of the leading Financial Advisory firm. They provide financial advisories across various critical domains such as accounting, taxation, commercial laws, banking laws, and treasury & risk management. They manage all the relevant documents for multiple countries as unstructured data.
Challenge:
The client was reviewing all the documents one by one to identify the rules and laws and then providing the advisory based on the information retrieved. The need was to leverage AI to enhance knowledge advisory across specific domains to empower organizations with the tools and insights needed to fast-track decision-making at all levels through comprehensive training on underlying rules, laws, frameworks, and use case references – An important assistant to CFO and Corporate legal teams. The greatest challenge for Chimera was to identify the right Foundation Model and the right RAG techniques that would fulfil the expectations.
Our Strategy:
Chimera used Knowledge Base features and multiple foundation Models – Query classifications and Response creations – through the Amazon Bedrock Services along with advanced prompt engineering techniques and custom RAG approaches. The user question along with the rules and output structure is injected into the model to generate the advisories. References to the actual documents are also provided as citations as links to download the respective and specific page as pdf file.
Our Solution:
The relevant data files are uploaded into S3 Bucket. These files are chunked and broken into embeddings that are stored into Atlas vector database. To optimize the performance in orchestrating the embeddings, the country, domain and certain other details are embedded as Metadata into each of the chunk. These vectors serve as the Knowledge Base for the generative AI Foundation Model to produce the output. Also, a Lambda function has been created to automate this process of chunking the data into vectors whenever a new file is uploaded into the S3 Bucket. The entire analysis and development from Scratch was completed by Chimera in 8 weeks.
Business Outcome:
- Users can use the platform to leverage AI to provide precise, actionable advice, making complex rules and standards easier to understand and apply.
- This implementation could bridge the gap between theoretical knowledge and practical application, helping organizations make informed decisions quickly and confidently.
- Helps Business users to expand their knowledge and support professional growth and decision-making capabilities.