CoinCompass — personal budgeting on Alexa+

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for CoinCompass — personal budgeting on Alexa+

Inspiration

We noticed that many people find traditional budgeting apps tedious and difficult to update on the go. We wanted to create a solution that makes financial tracking as effortless as having a conversation. By focusing on Alexa+, we aimed to show how generative AI can transform how we manage personal money.

What it does

CoinCompass allows users to track their spending by simply speaking to their Alexa device. It categorizes transactions automatically, monitors budget limits, and provides proactive advice based on historical data. The system acts as a virtual financial advisor that lives within the Alexa+ ecosystem.

How we built it

We built the backend using TypeScript and hosted our logic on AWS Lambda for scalability. We integrated Amazon Bedrock to handle natural language understanding and used the strands-agents-sdk to manage complex conversational flows. The core budgeting logic is managed through a custom framework integrated with Amazon Alexa.

Challenges we ran into

We initially struggled with mapping vague user phrases to specific financial categories in our budget logic. We also faced challenges in ensuring the latency of the Bedrock calls remained low enough for a seamless Alexa experience. Refining the agent prompts to handle various financial scenarios required significant iteration and testing.

Accomplishments that we're proud of

We are very proud of how accurately the Bedrock models interpret complex financial queries. We successfully integrated the strands-agents-sdk to create a fluid dialogue that feels natural and human. Seeing CoinCompass respond fluently on Alexa+ was a major win for our team.

What we learned

We learned a lot about orchestrating stateful AI agents within a serverless environment. We also gained deep insights into how to optimize TypeScript code for high-performance Lambda functions. This project deepened our understanding of the intersection between generative AI and conversational interfaces.

What's next for CoinCompass

We plan to integrate real-time bank API connectors to automate transaction fetching entirely. We also want to add predictive analytics features that warn users about potential overspending before it happens. Expanding the support for multiple languages on Alexa+ is also a priority on our roadmap.

Built With

• amazon-alexa
• amazon-bedrock
• aws-lambda
• budget
• coincompass
• strands-agents-sdk