Inspiration
The purpose of this project is to allow users to get quality customer support while elevating human agents from a constant stream of support queries allowing them to focus on client who need their help.
What it does
Engage AI is a customer assistant agent with a human in the loop for escalations. It includes a scheduler to notify the user of upcoming repayments through their selected channel of communication (SMS, WhatsApp or RCS if enabled). Responds to user queries based on training data and policy documents as its knowledge base and escalates questions outside the scope of its instructions and those not included in the policy or sample data.
How we built it
The User Interface is a React Application, built and deployed in an S3 bucket, users are placed in two Cognito pools (customer and agent) and lambda functions to bridge between the frontend, agent, Dynamo DB and Cognito pool . The Agent is built using python and has its own separate lambda functions to handle user communications, conversations fetching and escalations.
Challenges we ran into
Restricted resources in deploying the agent on our AWS lab due to bedrock-service constraints
Accomplishments that we're proud of
Seeing the system work as a whole, including the secure frontend behind the CloudFront URL, Agent and Customer login, and triggering the step functions to demonstrate user response and conversations being save in the DynamoDB.
What we learned
Collaboration across multiple vectors - Developers, AWS Cloud Engineers, and AI & Data Engineers
What's next for Engage AI
Building a production ready systems with real triggers to customer devices, capable of integration with different financial institution systems.
Log in or sign up for Devpost to join the conversation.