Executive summary
Client background
Our client is a rapidly growing brokerage firm with decades of market experience. With a substantial presence, the company manages assets totaling billions of dollars under administration, observing the annual opening of thousands of accounts. The firm provides low-fee trading solutions, positioning itself as a viable alternative to brokerage firms affiliated with major banks.
Business challenge
The client needed to streamline the high amount of routine tasks that were slowing down the work of their employees. They aimed to introduce generative AI to boost employee efficiency and productivity.
Value delivered
Sparknet helped the client develop a custom solution powered by generative AI. The solution boosted employee efficiency by streamlining tasks, as well as by serving as a go-to destination for obtaining quick and accurate information about the company’s policies, services, etc.
Success story in detail
Business challenge: boosting efficiency by streamlining manual tasks
The client wanted to improve the efficiency of their employees by allowing them to perform their routine tasks faster and easier. To achieve this, the company wanted to implement an internal web portal powered by generative AI that would streamline writing emails, creating JIRA tickets, describing application features, etc. The solution would also serve as a go-to source of information about the company regarding security, internal policies, etc. As a company that works in finance, the client needed to build a custom AI solution instead of using open-source AI tools.
Implementation: designing and implementing a custom web portal powered by generative AI
Sparknet designed the architecture for the internal web portal powered by generative AI. The platform helps employees to be faster and more efficient at performing various tasks, such as creating JIRA tickets with detailed descriptions, describing application features, and more.
We have used Single Sign-On based on MS Azure Active Directory to authenticate users. We have also used a custom .NET-based API Gateway as an entry point on the backend to aggregate user requests and orchestrate business workflows.
In addition, we have implemented an internal multi-tenant data storage. The data storage comprised all the information from the corporate website and other internal documents, such as an internal knowledge base, security manuals, Confluence, etc. This allows employees to receive quick and accurate answers to any company-related questions easily. Each time a user submits a question in the chatbot, the system receives the user’s question and the context from the corporate knowledge database. The text files from the knowledge base are indexed and only the relevant ones to the question are selected. Finally, a custom generative AI API developed by Sparknet processes the indexed context and sends a short and concise answer in a chatbot.
Furthermore, we incorporated MLOps practices, focusing on machine learning lifecycle management, and integrated support for Large Language Models (LLM) into the client’s operations.
To ensure data security, we implemented two levels of data protection through API requests to handle authentication and authorization and internal identity access management. We also implemented data encryption, a VPN connection, and a network firewall.
Finally, to reduce infrastructure expenses, Sparknet has explored the best ways of using GPUs and analyzed multiple large language models that best fit the client’s solution.
Value delivered by Sparknet streamlining operations and improving employee efficiency
Sparknet designed an internal web portal using generative AI to accelerate routine tasks. We also implemented a multi-tenant data storage that comprises information from the corporate website and internal documents. It enabled employees to receive answers to company-related questions. As a result, we helped the client improve their business in several significant ways:
- Improved employee efficiency by streamlining the completion of a wide range of repetitive tasks (emails, creating JIRA tickets, etc.);
- Helped users find relevant internal information faster by implementing an internal knowledge database.
