SignalZero

Inspiration

Financial fraud can be difficult to recognize when transactions are scattered across accounts and suspicious patterns are not immediately obvious. A large purchase, repeated charges, or a sudden burst of transactions might deserve attention, but simply flagging activity does not explain what happened or what someone should do next.

We wanted to build a tool that connects suspicious-activity detection with an understandable investigation process. That idea became SignalZero: a financial monitoring workspace that helps users move from a potential warning sign to a documented review.

What it does

SignalZero is a financial activity monitoring and investigation dashboard.

Overview: Summarizes analyzed transactions, suspicious alerts, open investigations, resolved cases, and activity trends.

Alerts: Organizes potential suspicious activity by detection type, severity, and status.

Investigations: Displays transaction details, supporting evidence, related activity, and spending-baseline context when sufficient history is available. Reviewers can record feedback and update case statuses.

Reports: Generates printable investigation records that include transaction details, detection reasons, evidence, status, and recorded feedback.

Accounts: Supports a Nessie banking-sandbox workflow using simulated financial data. A Plaid bank-linking integration is also in development.

SignalZero is designed to support human review, not automatically declare that a transaction is fraudulent. An alert is a reason to investigate, not proof of wrongdoing.

How we built it

We built SignalZero as a web application using React and TypeScript for the frontend, with Supabase for authentication, database storage, and backend Edge Functions.

We integrated the Capital One Nessie API to work with simulated banking data and designed the application around a connected workflow: transaction data, detection signals, investigations, and reports.

The interface includes dashboard metrics, transaction-activity visualizations, alert queues, investigation evidence, reviewer feedback, case-status controls, and printable reports. We also began integrating Plaid to explore a bank-connection workflow.

Challenges we ran into

One of our biggest challenges was connecting the frontend, authentication, database, and external financial APIs into a reliable end-to-end system within a limited hackathon timeframe.

We encountered issues with Supabase Edge Functions and the Plaid token-exchange flow, which required us to investigate server-side configuration, authentication, database tables, and API responses. We also had to think carefully about how to explain suspicious activity without presenting a detection signal as a confirmed fraud case.

These challenges reinforced the importance of testing integrations independently, handling errors clearly, and protecting sensitive credentials on the server rather than exposing them in frontend code.

Accomplishments that we're proud of

We are proud to have created a cohesive investigation workflow rather than just a page that displays transaction data.

SignalZero connects dashboard-level monitoring with individual alerts, supporting evidence, reviewer feedback, investigation statuses, and printable reports. We also built the experience around explainability and human oversight, so users can examine why an alert exists instead of blindly trusting a warning.

Building the prototype during a short hackathon pushed us to prioritize the features that make the product useful as a whole.

What we learned

We learned how many pieces must work together to build a financial application: frontend state, authentication, database permissions, backend functions, external APIs, and error handling.

We also learned that a useful fraud-monitoring tool needs more than detection rules. It needs to communicate uncertainty, provide understandable evidence, protect sensitive data, and give people a clear way to review and document findings.

What's next for SignalZero

Our next steps are to finish and test the Plaid connection flow, complete reliable transaction synchronization and database persistence, and strengthen the end-to-end experience.

We also want to improve detection rules, test them against more varied transaction scenarios, expand related-activity analysis, and make reports more useful for follow-up reviews. Longer term, we want to explore carefully evaluated risk-scoring methods while preserving transparency, privacy, and human oversight.

Built With

• api
• apis
• capital
• css
• financial
• html
• javascript
• nessie
• one
• plaid
• postgresql
• react
• rest
• supabase
• typescript
• vite