In the financial services sector, portfolio managers and research analysts constantly sift through vast amounts of data to gain a competitive advantage in their investments. According to the NVIDIA Technology Blog, the ability to make informed decisions depends on access to relevant data and the ability to quickly synthesize and interpret it.
Traditional Analytics vs. AI-based analytics
Traditionally, sell-side analysts and basic portfolio managers focused on a limited number of companies, meticulously reviewing financial statements, earnings calls, and corporate filings. Systematic analysis of financial documents across the broader trading universe was a difficult task, usually accessible only to sophisticated quantitative trading firms due to its technical and algorithmic complexity.
Traditional natural language processing (NLP) methods such as word corpora, sentiment dictionaries, and word statistics often fall short of the capabilities of large-scale language models (LLMs) in financial NLP tasks. LLMs have shown excellent performance in areas such as medical document comprehension, news article summarization, and legal document retrieval.
Enhanced features with NVIDIA NIM
By leveraging AI and NVIDIA technology, sell-side analysts, fundamental traders, and retail traders can significantly accelerate their research workflows, extract more nuanced insights from financial documents, and cover more companies and industries. By adopting these advanced AI tools, the financial services sector can enhance its data analytics capabilities, saving time and improving the accuracy of investment decisions. According to the NVIDIA 2024 State of AI in Financial Services survey report, 37% of respondents are exploring generative AI and LLM for report generation, synthesis, and investment research to reduce repetitive manual tasks.
Revenue Call History Analysis Using NIM
Earnings calls are a valuable source of information for investors and analysts. By analyzing these records, investors can gain valuable insight into the company’s future earnings and valuation. NVIDIA NIM provides tools to perform this analysis efficiently and accurately.
Step-by-step demo
The demo uses NASDAQ earnings release records from 2016 to 2020. The dataset contains a subset of 10 companies, and 63 records were manually annotated for evaluation. The analysis involves answering questions about revenue streams, cost components, capital expenditures, dividends or share buybacks, and material risks mentioned in the report card.
NVIDIA NIM Microservices
NVIDIA NIM provides optimized inference microservices for deploying large-scale AI models. Supporting a wide range of AI models, NIM leverages industry-standard APIs to ensure seamless, scalable AI inference on-premises or in the cloud. Microservices can be deployed with a single command, making them easy to integrate into enterprise-grade AI applications.
Building a RAG pipeline
Retrieval Augmented Generation (RAG) combines document retrieval and text generation to improve language models. This process includes document vectorization, query injection, document re-ranking, and answer generation using LLM. This method improves the accuracy and relevance of the information retrieved.
Evaluation and Performance
Performance evaluation of the search phase involves comparing the actual JSON with the predicted JSON. Metrics such as recall, precision, and F1-score are used to measure accuracy. For example, the Llama 3 70B model achieved an F1-score of 84.4%, demonstrating its effectiveness in extracting information from revenue call records.
Implications for financial services
NVIDIA NIM technology is poised to revolutionize financial data analytics. It will allow portfolio managers to quickly synthesize insights from a multitude of earnings calls to improve investment strategies and outcomes. In the insurance industry, AI assistants can analyze financial health and risk factors in company reports to improve underwriting and risk assessment processes. Banks can analyze earnings calls to assess the financial stability of potential loan recipients.
Ultimately, this technology will provide users with a competitive edge in their markets by improving efficiency, accuracy, and data-driven decision-making capabilities. Visit the NVIDIA API Catalog to explore available NIMs and experiment with LangChain integrations.
Image source: Shutterstock