FinMate

Just started your career? FinMate is your AI-powered advisor for choosing healthcare, savings, and retirement benefits, all personalized to help you achieve your financial goals.

Comment

Inspiration

When new employees begin their careers, one of the first major tasks they face is selecting their benefits package. For many, especially recent college graduates with little financial experience, this process can feel overwhelming and intimidating. Faced with unfamiliar terminology and long-term implications, employees often fear making the wrong choices.

We created FinMate to solve that problem. FinMate is an AI-powered financial assistant that guides employees through benefits selection, helping them feel confident, informed, and empowered during an otherwise confusing process. Its goal is not only to recommend the right benefits but also to teach users about financial decision-making and foster long-term financial literacy.

What it does

FinMate starts by learning about each employee through a personalized survey that captures their goals, priorities, and personal situation. Using this information, FinMate generates tailored recommendations on which benefits and financial goals to prioritize.

Employees can also chat with FinMate, powered by Claude, to ask questions in natural language and receive clear explanations of the recommendations. This chat feature empowers users to understand the “why” behind each choice, helping them make informed decisions and gain confidence in managing their own finances.

How we built it

Our frontend is built with TypeScript, Tailwind CSS, and Framer Motion, providing a smooth and intuitive user experience with clean animations. On the backend, we use Node.js running on multiple AWS Lambda microservices, orchestrated through API Gateway.

User data is stored in DynamoDB and queried using PartiQL, allowing for flexible and efficient access. For intelligent recommendations and personalized guidance, we integrated Claude Sonnet 4 through AWS Bedrock. To enhance the model’s context-awareness, we created a custom vector database containing benefits-specific information, financial guidance, and carefully engineered prompts to ensure responses remain accurate and helpful.

Challenges we ran into

We faced multiple infrastructure challenges during development. AWS repeatedly blocked our newly created accounts, causing Bedrock access to fail unpredictably. We eventually had to create multiple accounts and work closely with AWS engineers before obtaining a stable IAM account through codeLinc.

Additionally, MySQL Workbench failed to connect to our AWS RDS instance due to Lincoln Financial Group’s firewall. This prevented direct database access and forced us to explore alternative approaches. These challenges were frustrating but taught us resilience, creative problem-solving, and a deeper understanding of cloud services.

Accomplishments that we're proud of

We are proud of successfully connecting DynamoDB, API Gateway, and Lambda to implement full CRUD functionality. Designing this system on paper and seeing it function seamlessly in practice was extremely rewarding.

We are also proud of our work training the LLM. By creating a vector database of benefit information, financial guidance, and prompts, we enabled Claude to perform exceptionally well as a financial advisor. The model consistently provides thoughtful, context-aware responses that feel natural, respectful, and trustworthy.

What we learned

This project was our first hands-on experience with AWS, and we learned about serverless architecture, database design, security groups, IAM roles, and real-world cloud security practices. We also gained insight into agentic AI system design, including embeddings, vector databases, and the role of the LLM in an agentic workflow.

Mentorship from Lincoln Financial engineers helped us understand how to structure AI agents effectively. We applied these principles directly to FinMate, improving both the AI’s functionality and the overall user experience.

What's next for FinMate

FinMate already allows employees to export chat transcripts and benefit selections with the click of a button. In the future, we plan to integrate these features with HR systems via live API calls to automate actions directly.

We also aim to implement JWT-based authentication through AWS Cognito to provide a smoother and more secure login experience. Finally, we hope to add an AWS RDS instance for structured profile data while continuing to use DynamoDB for flexible storage, optimizing our data strategy and improving scalability.

https://drive.google.com/file/d/1hkhcIgm4Zcieez5crlFHQOdEZvJp7dqg/view?usp=sharing

Built With

Share this project:

Updates