Intelligent Document Processing

In financial services, processing documents involves complex data, such as loan records, external regulatory filings, transaction records, public market filings, and more. The sheer volume of this financial data makes it impractical to digest and extract insights manually, and current automated solutions for unstructured data processing are inefficient.

Digitization and automation can streamline operations, reduce errors, and help organizations stay competitive in banking and insurance. In capital markets, AI algorithms can automate the processes of analyzing market trends, identifying patterns, and executing trades, which reduces time to action.

But individuals still face the time-intensive task of manually inputting data from paper-based documents. Through the integration of AI, users can seamlessly coordinate a diverse range of services and harness machine learning capabilities with intelligent document processing (IDP).

Machine learning models can identify a wide variety of document types and extract relevant information from them. With generative AI, organizations can summarize structured and unstructured documents for research analysts, loan processors, and customer service agents.

For capital markets, the most powerful investment insights are hidden in unstructured text data from everyday sources such as news articles, blogs, and SEC filings. Generative AI lets traders provide deeper insights from unstructured data than traditional tabular data analysis, enabling faster decision-making and reducing the risk of financial losses.

Financial institutions are developing deep learning algorithms to automatically process documents digitized by their clients for various financial products. In retail banking for real estate, machine learning models accelerate title search, underwriting, and closing processes for home loan documents, helping complete home transactions much faster than before. With natural language understanding (NLU), banks can rapidly interpret the numerous requests and inquiries that occur during the due diligence process for loans and transactions.

NVIDIA provides resources for financial institutions looking to use generative AI for IDP, such as constructing chatbots with retrieval-augmented generation (RAG) to automate loan processes or developing market insights in portfolio construction and trade execution.

Optimal Inference for Generative AI Workloads

NVIDIA NIM, part of NVIDIA AI Enterprise, is a set of easy-to-use inference microservices designed to accelerate the deployment of generative AI across your enterprise. This versatile runtime supports open community models and NVIDIA AI Foundation models from the NVIDIA API catalog, as well as custom AI models. NIM builds on NVIDIA Triton™ Inference Server, a powerful and scalable open-source platform for deploying AI models, and is optimized for large language model (LLM) inference on NVIDIA GPUs with NVIDIA® TensorRT™-LLM. NIM is engineered to facilitate seamless AI inferencing with high throughput and low latency, while preserving the accuracy of predictions. NIM lets organizations deploy AI applications anywhere with confidence, whether on premises or in the cloud. 

Accelerate Data Curation

NVIDIA NeMo™ Curator is a scalable, GPU-accelerated data-curation microservice that prepares high-quality datasets for pretraining and customizing generative AI models. With it, financial institutions can train and fine-tune LLMs on financial documents. NeMo Curator streamlines data-curation tasks such as data download, text extraction, reformatting, cleaning, quality filtering, and exact/fuzzy deduplication to help reduce the burden of combing through unstructured data sources. Document-level deduplication ensures that LLMs are trained on unique documents, which can greatly reduce pretraining costs.

Real-Time Information Retrieval

NeMo Retriever is a collection of CUDA-X™ microservices that enable semantic search of enterprise data to deliver highly accurate responses using retrieval augmentation. Developers can use these GPU-accelerated microservices for specific tasks, such as searching for relevant pieces of information within internal data to answer business questions, increasing accuracy and reducing hallucinations.