SmartRetain AI - Customer Churn Intelligence Agent

• 

  SmartRetain-AI Streamlit interface showing customer data input and churn prediction workflow.

• 

  AI-generated churn risk prediction with personalized retention strategy recommendations.

• 

  Estimated business impact analysis and AI-generated personalized customer communication for retention.

Inspiration: Customer churn is a challenge that almost every business faces, especially companies that depend on subscriptions or long-term customer relationships. While many tools can predict churn, they often stop there they don’t help businesses understand what actions to take next. I wanted to build something more practical. The idea behind SmartRetain-AI was to create a system that not only predicts which customers are likely to leave but also suggests personalized strategies to retain them using AI. In simple terms, I wanted to connect data science with real business decisions.

What it does: SmartRetain-AI is an AI-powered platform that helps businesses identify customers who might churn and provides recommendations on how to retain them. The system: • Predicts churn probability using machine learning • Categorizes customers into Low, Medium, or High risk • Generates personalized retention strategies using Amazon Nova (LLM) • Provides actionable insights businesses can use • Allows downloading results for further analysis Instead of just showing predictions, the platform focuses on actionable intelligence.

How I built it I built this project by combining traditional machine learning with generative AI. First, I trained a churn prediction model using the Telco Customer Churn dataset. Then I connected the prediction results with Amazon Bedrock’s Nova LLM to generate retention strategies based on the customer’s risk level.

I used: • Python and Scikit-Learn for the ML model • Pandas for data processing • Amazon Bedrock (Nova) for AI strategy generation • Streamlit for the user interface My goal was to keep everything simple, interactive, and practical.

Challenges I ran into: One of the main challenges was making sure the AI recommendations actually aligned with the churn predictions. This required experimenting with prompts and improving how model outputs were passed to the LLM. I also faced issues related to API security and environment configuration, which helped me understand the importance of protecting credentials when working with cloud services.

What I learned: This project helped me understand how different AI technologies can work together in a real-world application. I learned: • How to build an end-to-end ML pipeline • How to integrate LLMs into business workflows • Prompt engineering for meaningful outputs • Deploying interactive apps with Streamlit • Practical cloud integration using AWS Most importantly, I learned how to turn data into decisions, not just predictions.

What’s next for SmartRetain-AI I would like to expand this project further by adding: • Real-time data integration • Dashboard visualizations • CRM system integration • Automated marketing recommendations • Cloud deployment for scalability My long-term vision is to develop this into a complete customer intelligence tool for businesses.

Built With

• bedrock
• generative-ai
• llm
• machine-learning
• nova
• pandas
• python
• scikit-learn
• streamlit