Aegis

Our Story:

What Inspired Us: We were interested in the potential of Web3, but we also noticed how intimidating crypto payments can be for people who are not familiar with blockchain technology. A simple payment can involve wallet addresses, networks, transaction approvals, and other technical details that can easily confuse a new user. At the same time, users may not always realise when a transaction looks unusual, such as sending a much larger amount than normal or transferring funds to a new recipient. This led us to ask a simple question: What if interacting with crypto payments could feel more like having a conversation with an assistant, while still giving users an extra layer of protection? That idea became Aegis.

What We Learned: One of the biggest things we learned was that building a Web3 application is not only about getting a blockchain transaction to work. The experience around that transaction matters just as much. We learned how conversational AI can be used to turn natural-language instructions into structured payment requests. We also explored how machine learning can identify unusual transaction patterns and how blockchain can provide a transparent record of completed transactions. More importantly, we learned that these technologies become much more useful when they are connected together to solve an actual user problem.

How We Built Aegis: We designed Aegis around three main layers. First, the AI chat layer allows users to describe what they want to do in natural language. For example, a user can type: "Pay Alex 50 USDC every Friday." The system interprets the request and turns it into structured payment information. Next, the fraud-detection layer checks transactions for unusual behaviour. We created a synthetic dataset of fictional transactions containing both normal and unusual activity. The dataset includes information such as transaction amount, recipient, and timestamp. Examples of unusual behaviour include unusually large payments, new recipients, and transactions occurring at unusual times. Finally, the blockchain layer allows approved payments to be executed on a public testnet. This lets us demonstrate that the process goes beyond a simulated chatbot response and can result in an actual on-chain test transaction that can be verified through a block explorer.

The overall workflow is: Chat → Understand → Check → Explain → Approve → Execute

Challenges We Faced: One of our biggest challenges was bringing several different technologies together into one simple experience. AI, machine learning, and blockchain each have their own requirements and behave very differently. Making them work together while keeping the experience understandable for a normal user required a lot of testing and iteration. Another challenge was working with limited time. Instead of trying to build every possible feature, we focused on demonstrating the core idea clearly: a user should be able to describe a payment, have Aegis assess it for unusual behaviour, understand the warning, and then see the approved transaction executed on the blockchain testnet. We also had to think carefully about how to present risk detection. Simply telling a user that a transaction was "suspicious" isn't very helpful. We wanted Aegis to explain why something was flagged so the user could make a more informed decision.

What We Hope to Do Next: Aegis is currently a prototype demonstrating the concept of an AI-assisted crypto payment guardian. In the future, we would like to make the system more accurate using larger and more diverse transaction datasets, improve the explanations provided by the AI, support more complex payment workflows, and explore additional security mechanisms. Our goal is to make Web3 easier to approach without hiding the technology that makes it possible. Aegis is our first step toward making crypto payments feel less complicated and more understandable — while helping users make safer decisions before their money moves on-chain.