Inspiration
Online payment scams, especially Authorized Push Payment (APP) fraud, can trick people into willingly sending money to scammers. The payment may look legitimate, but the person or situation behind it could be suspicious.
I wanted to explore a simple but important question: What if people had one more chance to think before sending money?
That question inspired me to build Pause — a human-first payment safety checkpoint designed to help people recognize potential risks before a payment proceeds.
What it does
Pause evaluates payment details and provides one of three decisions:
- ALLOW: No significant risk indicators were identified.
- HOLD: The payment needs additional attention or verification.
- BLOCK: Strong risk indicators suggest that the payment should not proceed.
It combines rule-based risk checks with an optional AI analyst powered by the Vultr Inference API. The AI analyst can recommend a higher risk level, but it cannot lower the severity determined by the existing checks.
Pause also integrates with the PayPal Sandbox to demonstrate payment authorization, capture, and void operations in a controlled environment.
The goal is not to replace human judgment, but to give people useful information before they make a potentially irreversible decision.
How I built it
I built Pause using Java 21 and Spring Boot for the backend, with HTML, CSS, and JavaScript for the frontend.
Technology stack:
- Java 21
- Spring Boot
- Maven
- H2 Database
- HTML, CSS, and JavaScript
- PayPal REST API (Sandbox)
- Vultr Inference API
I designed the backend to handle payment assessment, risk evaluation, and payment operations. The frontend allows users to review payment details and understand the resulting safety decision.
I implemented rule-based checks as the foundation of the risk assessment system and added an optional Vultr AI analyst for additional analysis. I also designed a safety constraint so that AI recommendations can increase risk severity but cannot reduce it.
PayPal Sandbox provides a controlled way to demonstrate the payment lifecycle without using real money.
Challenges I ran into
One of my biggest challenges was designing how traditional rule-based checks and AI recommendations should work together.
I did not want an AI model to make unrestricted payment decisions. Instead, I designed the system so that deterministic checks remain the foundation, while AI recommendations can only increase the level of caution.
Integrating the PayPal Sandbox payment lifecycle, configuring external API credentials, and connecting the frontend with the backend also required careful testing.
Since I was relatively new to Spring Boot, learning its application structure, controllers, services, configuration, and integrations while building a complete project was another challenge.
As a solo developer, managing development, debugging, testing, and integration within a limited hackathon timeline pushed me to make careful technical decisions and prioritize the most important features.
Accomplishments that I'm proud of
- Independently designed and built a payment safety prototype from the ground up.
- Integrated PayPal Sandbox authorization, capture, and void operations.
- Implemented a three-level risk decision system: ALLOW, HOLD, and BLOCK.
- Integrated an optional Vultr-powered AI analyst with a safety constraint that prevents it from lowering risk severity.
- Designed a user interface that makes payment risk decisions easier to understand.
- Gained practical experience building and integrating a Java Spring Boot application.
Most importantly, I'm proud that I turned a simple idea into a working prototype centered on one principle: payment safety should involve informed human decisions, not blind trust in automation.
What I learned
Building Pause taught me that integrating AI into a financial workflow is not just about generating predictions. It is equally important to define the boundaries of what AI is allowed to do.
I gained hands-on experience with Spring Boot, REST APIs, PayPal Sandbox integration, database configuration, external AI services, backend testing, and application deployment.
I also learned the importance of transparent risk decisions. A system should communicate uncertainty rather than presenting an AI recommendation as a guarantee that a payment is safe.
Working independently on this project also strengthened my problem-solving, debugging, and time-management skills.
What's next for Pause
Pause is a prototype with plenty of room to grow. My future goals include:
- Adding user authentication and personalized payment histories.
- Implementing trusted-contact verification and notifications.
- Exploring webhook-based payment event handling.
- Improving risk detection with additional scam indicators.
- Strengthening testing and deployment security.
- Making the payment review experience more accessible and understandable.
My long-term vision is to make Pause a reliable safety layer between a user's intention to pay and the actual movement of money.
Pause. Think twice. Pay once.
