Puran

You know when you doomscroll for 2 hours and regret 80% of it? We prevent that regret from how you spend your money! Puran sets near-term financial goals to help you spend on the things that you want.

Comment

Inspiration

We have all doomscrolled before. You pick up your phone for a second and suddenly it has been an hour. The first few minutes feel fine, but the rest just feels like lost time. There are apps that help you fix that with your time. Nobody has built that for your money. Americans spend over $282 on impulse buying every month (Capital One Shopping)! We built Puran to bridge that gap. Instead of worrying about a retirement 40 years away, we wanted to help people skip two Chipotle runs this week to afford a concert with friends in three.

What it does

Puran is a short-term financial route-planner that uses gamification to build better spending habits.

  1. The Ingestion: Users upload their transaction history (CSV/PDF) to an Amazon S3 bucket. We then create graphs to show and compare your previous month's spending, your biggest spending categories etc..
  2. The Intelligence: Amazon Bedrock analyzes that data to find "spikes" and patterns, like that 3x-a-week takeout habit. We create 3 recommendations from this analysis which you can either accept or reject. We also project your potential monthly savings from our suggestions.
  3. The Goal: Users set "Micro-goals" (less than a month long) that are monetary and measurable.
  4. The Simulation: Your progress is visualized as a forest. Consistent saving grows trees; overspending or missing goals causes "burn events" that erode your digital ecosystem.

How we built it

We leaned heavily into a serverless AWS architecture to keep the app fast and secure

  • Frontend: Built with React and hosted via AWS Amplify for seamless deployments.
  • Backend: We used AWS Lambda (Over 10 lambda functions) and API Gateway (Custom REST API) to handle logic without managing servers.
  • Data Pipeline: Files land in Amazon S3, triggering a Lambda function that strips personal info and pushes clean data into Amazon DynamoDB.
  • AI Layer: Amazon Bedrock acts as our pattern detection engine, suggesting specific spending cuts to meet goal deadlines.
  • Security: All service interactions are governed by strict AWS IAM roles to ensure data privacy. We manually edited the policies for each lambda.

Challenges we ran into

Figuring out what services to use was very hard. We decided on API Gateway with a multiple lambda services. We then created custom triggers for each lambda, and setup policies to prevent any excessive usage. We also failed a lot initially to connect with AWS, so we used CloudWatch logs and our console to debug what parts were failing.

Accomplishments that we're proud of

All eight AWS services working together in a live demo. The pipeline runs end to end. The forest responds to goal state in real time. We are proud that it actually works and that it feels like something a real person would use.

What we learned

AWS diagrams look simple. The actual wiring is not. We spent way more time on IAM and event triggers than expected. We also learned to cut early: dropping authentication and bank integration from scope early meant the core product got built properly instead of everything getting half-finished.

What's next for Puran

Authentication through Cognito would be the next logical step that can be added to this. We would also switch to an Agentic structure rather than just pull from AWS Bedrock.

Share this project:

Updates