Revolutionizing Banking Operation with AWS Generative AI Services
Gautam Nadkarni, Chief Technologist, Wipro Full Stride Cloud Services
Introduction
The banking industry is undergoing rapid transformation, driven by the need to innovate beyond traditional practices. Generative Artificial Intelligence (Gen AI) emerges as a pivotal technology in this evolution, with Amazon Web Services (AWS) at the forefront, offering advanced Generative AI tools. Generative AI differs from conventional AI by its ability to generate new content or insights from existing data, mimicking human creativity.
These Gen AI tools are enabling banks to revolutionize operations, enhance customer experiences, and bolster risk management.
GenAI Opportunity Mapping for Banking Capabilities
Generative AI’s applications in banking are vast. It can generate artificial data for training machine learning models, addressing privacy concerns with actual customer data. It can also deliver personalized banking services by analyzing customers’ transaction records to offer tailored financial advice and product recommendations. Furthermore, it enhances security by creating synthetic identities for testing security measures without exposing real customer data. Key banking use cases of Gen AI include:
· Fraud Detection and Prevention: Leveraging AI to analyze transaction data for patterns indicating fraud, enabling real-time anomaly detection to protect customer assets. For instance, Wipro’s Fraud Detection solution, built on Amazon FD, has been implemented in a US bank.
· Personalized Customer Experiences: AI enables banks to offer customized recommendations and support, fostering stronger customer relationships and satisfaction.
· Risk Management: AI models analyze past data and market trends for informed risk assessment in lending and investments, aiding in better decision-making.
· Regulatory Compliance: AWS AI tools automate data analysis and reporting for compliance, reducing operational costs and compliance-related risks.
A broad mapping of AWS Generative AI opportunities across the banking business capabilities are depicted in the picture below:
AWS Generative AI in Banking:
AWS offers a suite of Generative AI tools tailored for the banking industry, enhancing operations through generative AI. A list of AWS Generative AI capabilities for banking industry are provided here — https://aws.amazon.com/financial-services/generative-ai/ . Key AWS Gen AI capabilities include Bedrock, Sagemaker, Amazon Q, PartyRock, Nvidia GPU-Powered EC2 instance, Trainium, Inferentia and EC2 Ultraclusters.
A good depiction and illustrative mapping of some of these AWS Gen AI services at infrastructure, tools and applications layers has been provided and described in an AWS AI/ML Blog here. The figure, from this blog, depicting the illustrative layered mapping of AWS Generative AI services is reproduced and provided below –
Typical Architecture for a Generative AI Application on AWS Bedrock:
While there are many ways in which generative AI applications can be architected on AWS, an illustrative technical architecture is depicted in the diagram below that adopts a retrieval augmented generation (RAG) approach to enable faster development of the Gen AI application. RAG approach requires lesser model training due to usage of existing Generative AI LLMs in conjuction with existing knowledge sources.
Key Architecture Components:
· Claude LLM: A generative large language model central to generating human-like text responses from input prompts and retrieved knowledge.
· Guardrails: Tools and practices ensuring content safety, regulatory compliance, and adherence to ethical guidelines by filtering harmful content.
· Prompt Engineering: The technique of crafting effective prompts to guide Claude LLM in generating relevant and high-quality outputs.
· Knowledge Chunking: The process of dividing the knowledge base into smaller, semantically coherent chunks for quicker retrieval.
· Titan Embeddings: Converts text into high-dimensional vectors for efficient semantic search within the knowledge base.
· OpenSearch Vector Database: Utilizes AWS OpenSearch Service with vector search capabilities for storing and retrieving Titan Embeddings.
· S3 Buckets: Amazon Simple Storage Service (S3) stores the knowledge base and prompt store, containing data for retrieval and crafted prompts for Claude LLM interaction.
Architecture Overview:
· Knowledge Base and S3: The knowledge base is stored in S3 buckets, where data is chunked and processed to generate Titan Embeddings for efficient retrieval.
· OpenSearch Vector Database: Indexes Titan Embeddings, enabling semantic search queries against the knowledge base to find relevant information swiftly.
· Claude LLM and Prompt Engineering: Crafted prompts stored in S3 buckets refine interactions with Claude LLM, ensuring accurate understanding and response generation.
· Guardrails: Content generated by Claude LLM is vetted for compliance, safety, and ethical standards before presentation to the user.
· Retrieval-Augmented Generation: The system retrieves relevant knowledge chunks using semantic search, which, along with input prompts, guides Claude LLM in generating informed responses.
Scalability and Performance:
· AWS Auto Scaling: Ensures the architecture can efficiently handle varying loads by scaling resources as needed.
· Amazon CloudFront: Improves content delivery speed and reduces latency, particularly for static content from S3 buckets.
· AWS Lambda: Offers a serverless option for processing tasks like prompt engineering and knowledge chunking, scaling with demand cost-effectively.
Examples of Generative AI offers for Banking on AWS Marketplace:
Besides the AWS PaaS AI and Gen AI services, there are some of the generative AI solutions for banking sector that are available on AWS Marketplace. A few examples are listed below:
· InferQ — GenAI-powered identity provider (IDP) for banking
· Futurex — Cloud Payment HSM platform
· Q-Dox — AI-powered financial document processing
· Wipro AWS AI/ML Lab for banking use cases (Link)
Challenges and Considerations:
The use of AWS Generative AI in the banking sector has great innovation potential, but it comes with its own set of challenges; namely: data privacy, security, model interpretability, ethical considerations and integration with legacy systems.
The Future of Banking with AWS Generative AI:
AWS is constantly improving its Generative AI technology, which is expected to revolutionise the banking industry by improving efficiency, personalised services, and risk management. With AI insights, banks are better equipped to pre-empt customer needs, minimise risks, and foster growth in a competitive market. Gen AI is acting as a driving force for change, enabling banks to rethink their processes, enhance customer interactions, and discover fresh ways to generate value. As the banking sector welcomes this new approach, innovators will establish higher performance benchmarks and reshape the financial landscape.
About the author:
Gautam Nadkarni, Chief Technologist, Wipro Full Stride Cloud Services — leads the Cloud Transformation, Gen AI for IT and Innovation initiatives for the Asia Pacific region. He assists the Wipro customers in technology transformations and architecting innovative solutions for business problems.
