South African fintech consultancy firm Elenjical Solutions (ES) has developed three proof-of-concept AI apps focusing on using proprietary information for rapid information retrieval, code generation, and simplified interfacing with complex systems.
The machine learning apps use artificial intelligence to marshal large and diverse amounts of stored information from multiple sources as well as generating proprietary code from natural language inputs to interact with complex systems and data sources.
“Our goal is to reshape the traditional way our clients interact with data,” says Bereket Demeke, executive at ES. “In the complex landscape of financial markets our new apps are not just tools – they are methods enabling clients who work in fintech sector to engage with their data effortlessly.”
The apps aim to provide a cohesive, unified method for sourcing and synthesising information across financial companies’ complex channels. They hope to acheive substantial time savings by significantly reducing the hours spent on repetitive tasks, increasing automation, and as a result moving businesses towards greater operational efficiency and smarter resource management.
ES’s Information Retrieval App transforms access to internal knowledge bases, including SharePoint and Confluence libraries. Operating with a conversational AI interface, it enables company employees to work with information more efficiently.
In addition, the app can read and extract insights from various document formats such as PDFs, PowerPoints, and Word Docs.
“Unlike generic AI models, such as Chat GPT, our app understands domain specific information such as financial markets and capital markets technologies. That’s making it a must-have tool for financial professionals,” says Siju Mammen, technology lead at ES.
Natural Language to Proprietary Code App specifically targets the generation of proprietary code from natural language. The first use case of this application was tested for generation of the Murex MSL code. The first version of the app requires a developer to operationalise the code, but in upcoming editions, ES aims to evolve it into an agent model, enabling it to independently deploy, run, and test the code.
Natural Language to SQL App transforms natural language queries into SQL valid for a given database schema. It’s designed to handle structured data queries in complex financial databases.
The company is currently testing and improving these applications internally, and has started an AI consulting practice to introduce the services to the market.