DEV Community

Veronika
Veronika

Posted on Edited on

Building Revi: An AI-Powered Personal Economic Trainer at HackJunction 2026

What I Built

At HackJunction 2026 in Finland, my team, Project Aethon, developed Revi - a Personal Economic Trainer for S-Bank's challenge.

The idea was to help people understand their finances through natural conversations with their own financial data. Instead of simply showing transaction lists and charts, Revi allows users to ask questions like "Where did my money go?" or "What can I safely spend this week?"

Revi analyzes everyday spending, regular commitments, and necessary but irregular expenses, helping users understand where their money goes and make more informed decisions.

One of the features I find particularly valuable is financial scenario planning. For example, users can explore whether they could afford to buy a car next year, understand the impact on their budget, and test alternatives such as saving for three more months or adjusting their current spending.

We wanted to build more than another expense tracker. Our goal was to create an interactive financial coach that explains its reasoning and helps people understand the consequences of their decisions.

Demo

Try Revi: https://70.34.209.210.nip.io/

Demo video:
https://youtu.be/S73vF3zOar8
https://www.youtube.com/watch?v=G3yiPbq_qOk

Source code: https://github.com/victor06-git/project_aethon

Presentation: https://files.hackjunction.app/presentations/45/e9ac266d-6692-444f-b8ca-f412b93df2ff.pptx

Partner Technologies

We combined several partner technologies to bring Revi to life.

Exa API - Real-World Context

We integrated Exa's search technology to enrich financial conversations with relevant external information. It allows the trainer to retrieve current prices, available options, and open-data benchmarks, including information from Statistics Finland.

This was particularly useful for financial planning scenarios, where decisions depend not only on a user's transaction history but also on changing real-world conditions. With the user's permission, the trainer is designed to revisit agreed plans as external information or personal circumstances change.

Vultr - Backend Infrastructure

We used Vultr to host and deploy our backend infrastructure. This gave us a dedicated environment for running the application and handling financial conversations.

Google Gemini API & Anthropic Claude API

We integrated both APIs to power Revi's AI capabilities and natural-language interaction, allowing users to communicate with their financial data in a conversational way.

We also designed the architecture with future multi-bank integration in mind. Although PSD2 functionality was not active in the initial prototype, the system was structured to support incorporating account data from other banks in the future.

Hackathon Experience

Being selected to participate in HackJunction 2026 in Finland was an incredible experience. I was among 1817 selected participants from over 100 countries, surrounded by people with different backgrounds, skills, and perspectives.

Together with my teammates, Divin Joseph, Victor Asensio Bermúdez and Alejandro González Hernández, I worked on Revi under the pressure of a 48-hour deadline. We had to turn an idea into a working product, make technical decisions, integrate different technologies, and prepare our final presentation.

One of the most memorable moments was presenting our project to an audience of approximately 1817 people. It was exciting to see how much a team can accomplish in such a short time when everyone brings their own perspective and skills.

Beyond the technical experience, I will remember the people I met, the conversations, the atmosphere, and the opportunity to challenge myself in such an international environment.

HackJunction reminded me that some of the best learning happens when you step outside your comfort zone, work with new people, and build something together under real constraints

Top comments (0)