Sheryl

High Level Architecture diagram

Inspiration

I wanted to create an agentic AI that could perform the traditional tasks of a virtual assistant managing emails, scheduling meetings, and sending messages but in a more intelligent, conversational way. I’ve always been fascinated by how AI can automate everyday professional tasks and make work smoother.

What it does

Sheryl is an AI-powered virtual assistant built with Amazon Bedrock and Spring AI. She helps users manage emails, organize their calendar, and send WhatsApp messages all through natural conversation.

How we built it

Sheryl was built using Spring Boot as the core backend framework, integrated with Spring AI and Amazon Bedrock for conversational intelligence.
I used Bedrock Converse for multi-turn dialogue and connected APIs like Gmail and Google Calendar to make Sheryl truly functional as a productivity assistant.

Challenges we ran into

I had never worked with AWS AI services before, so understanding Bedrock’s setup and credentials was a big challenge.
Coming from a third-world country, I was also concerned about AWS costs, so I had to manage my usage efficiently.
Integrating the AI logic smoothly into Spring Boot without breaking configurations took a lot of trial and error.

Accomplishments that we're proud of

I successfully built a working AI assistant that integrates Bedrock and Spring AI something I had never done before.
Seeing Sheryl respond intelligently and interact with external services like Gmail and Calendar was a huge milestone.

What we learned

I learned how to integrate Spring AI with Bedrock, work with Bedrock Converse for chat-like experiences, and handle AWS configuration and credentials properly.
This project also taught me a lot about agentic architectures and how conversational AI can be embedded into real-world tools.