The Full Story
LLM Suite
The JPMC LLM Suite is a system of tools and technologies developed by JPMorgan Chase & Co. to enhance their capabilities in natural language processing (NLP) and machine learning through an AI Q&A chat service.
1. Natural Language Understanding: The suite includes tools that can understand and process human language, allowing the bank to analyze and interpret large volumes of text data.
2. Automation: It helps automate routine tasks by generating responses, processing documents, or handling customer inquiries with minimal human intervention.
3. Data Analysis: The suite can extract insights from unstructured data, such as emails, reports, and social media, to support decision-making and strategy.
4. Custom Solutions: JPMorgan Chase uses these technologies to build tailored solutions that address specific business needs and improve efficiency.
Generative UI
Generative UI is about using smart technology to automatically create and adjust designs for websites apps and systems. It can be a more interactive tool to view information such as charts, graphs, visuals, and graphics. However, it is based on the needs of a user and can also include personalized elements like recommended or suggested content of information.
LLM Suite: An Implementation of Generative UI
As an intern at JP Morgan Chase, I worked on the LLM Suite project, which involved implementing Generative UI elements such as graphs, visuals, and personalized content within an AI chat service. This initiative aimed to enhance versatility in AI output and improve the organization of large amounts of information. Due to compliance and data protection policies, I am unable to share specific prototype work or representations of LLM Suite. This overview provides a general insight into the work I contributed to during my internship.
Research
For the LLM Suite project, I conducted user interviews with 20 individuals from various fields, including design, business, trading, writing, risk analysis, product management, and software development. By synthesizing the results, I discovered that users prioritized clarity and conciseness in information distribution and question setup within the tool, while also valuing a streamlined process that wasn't overly time-consuming.
Additionally, users emphasized the importance of integrating citations or recommendations to enhance accuracy. To gain deeper insights into user perspectives on AI implementations, I created an empathy map that captured how users might think, feel, say, and act. The double diamond method of design thinking was utilized in the project process.
​
Market Analysis
As part of the LLM Suite project, I conducted a thorough competitor analysis of AI tools used by rival banks, evaluating their strengths and weaknesses. This analysis provided valuable insights into existing solutions, which I used to inform the development of the LLM Suite. Building on this research, I designed a flow for how generative UI elements could be integrated into the Suite tool, ensuring that our implementation would be both innovative and user-friendly.
Wireframing & Prototyping
I iterated through 10 different wireframes, both low-fidelity and high-fidelity, eventually creating a prototype that integrated various generative UI elements. For instance, when asked about the equity market from January to June 2024, the system would produce not just text but also graphs, widgets, flowcharts, and links to relevant articles. I also designed additional widgets, such as a Google Maps integration and global time widgets, and demonstrated how generative UI could be used to present Standard Operating Procedure processes through interactive forms and buttons.
