The client is a global Fintech firm offering hyper-personalized financial services.
The Challenge
As customers increasingly demand tailored financial services, retail banks must offer hyper-personalized product offerings at scale. Our client – a successful global Fintech – wanted to differentiate themselves from competitors by understanding customer spending habits and preferences and crafting personalized product and service recommendations. The company wanted a digital solution that would enable them to generate personalized marketing and product offers quickly and at scale. Technically, this called for a Generative AI-based analytical solution capable of identifying new product recommendation opportunities based on customer behaviors – and our client saw this approach as the key to engaging and retaining customers.
The Solution
Persistent developed a three-part solution designed to understand the client’s customers and target them intelligently for increased business. First, we built a large language model (LLM)-based Profile Generation Engine that produces natural language reports on user spending habits by analyzing transaction data. Second, we added a Product Recommendation Engine, leveraging LLM to generate personalized offers. And finally, we harnessed the power of Large Language Models (LLMs) to generate and edit persuasive, personalized emails to targeted customers.
AWS cloud and the AWS product set were key to our development effort. The team employed Amazon SageMaker to quickly and easily build and train machine learning (ML) models, and then directly deploy them into a production-ready cloud environment. We used AI21 Jurassic 2 Mid, a pre-trained LLM by AI21 Labs that can tackle complex tasks such as question response, summarization, copy generation, and advanced information extraction. We also leveraged some of the ready-to-use models in AWS Jumpstart to accelerate the development effort.
Other products and technologies employed in this innovative build included the LLAMA2-7B LLMs from Hugging Face (a collection of pre-trained and fine-tuned generative text models); LangChain’s use cases for rapid application creation; and a Pandas DataFrame (a Python library for working with data sets). The entire solution was coded in Python and React.js.
The Outcome
Persistent’s multi-phase GenAI/ML solution enabled the client to identify patterns from customer transaction data, generate profile summaries, and then use those summaries for personalized marketing. All the output from the solution can be described as “explainable recommendations,” — the LLM-generated personalized recommendations and marketing content are clearly based on known, observable customer behaviors. In short, GenAI-driven data insights now enable Fintech clients to transform transactional data into interpretable human profile summaries and attractive personalized product offers that drive new business.
Technology used:
- AWS Sagemaker & AI21
- Jurasic 2 Mid
- AWS Jumpstart
- Fine-tuned
- LLAMA2-7B
- LangChain
- Pandas
- Python
- React.js
