Conversational Search in App

Client

A leading fintech platform releasing an app for tracking personal finances

Industry

Fintech , Mobile Applications

Services

Helped in setting up ML team, 6 months roadmap for product, project execution in phased manner

Description of Project

A leading fintech client was developing a personal finance application that links directly to bank accounts to extract your bank statements.

1. Each item in the statement is intelligently parsed and tagged for expense categorization

2. Wanted to implement conversational search - “Show me expenses from last Friday” , “All purchases on Amazon in the last month” etc

3. Intelligent extraction of merchant

The client did not have previous AI expertise as it was their first large scale project. However, they had invested heavily in the system with an application launched in beta and already collecting millions of user transactions.

Solution

To decrease time to market and introduce intelligence within the application -

1. We developed a phased plan for 3 months, by breaking down the larger tech goals into iterative releases

2. Helped in identifying and hiring a initial ML team

3. Data strategy - identified the different types of data for each AI use case, helped in annotation strategy and a regular pipeline

Result

• Launched conversational search at 80% accuracy in 2 months

• Improved transaction categorization using a custom trained NLP model by 40%

• Leveraged Generative AI LLMs - OpenAI GPT and LLamav2 to speed up AI use cases like auto annotation and categorization