AthenaAutoTrader

Reliable trading bots for everyone

Comment

What it does

Athena AutoTrader is a webapp for building your own custom trading bots using a blockbased programming language. It's built to be intuitive and help more people learn the power of intelligent investing. We also integrate the new reasoning model Gemini 2.5 Flash to allow feedback about the trading bot you have built with nothing more than the press of a button while having a look at backtracking analysis with historical real market data.

Inspiration

We were inspired by the growth of codeless programming languages and interfaces. Scratch was a major inspiration and we wanted to provide a similar level of simplicity to a different audience. Where Scratch targets young programmers, we wanted to target financebros. After learning about Revoluts internal tool "Tower" we also took some inspiration with trying to keep the end user fully separate from the code running behind the scenes. We wanted something everyone would be able to use and understand, but with the potential for advanced behaviour through the use of fundamental logical operators.

How we built it

The webapp is composed of various different technologies. We chose Vite + React for out frontend, python and MongoDB in the backend for storing our data. We divided the team into primarily 1 frontend and 2 backend and 1 full stack, but collaboration was essential to integrate the different modules together in a seamless way. There were a lot of decisions that had to take place and we used a pros and cons discussion technique to collectively conclude on good solutions. The first evening we spent some time dividing tasks and setting a scope for the project which ended up being highly rewarding in terms of gained efficiency later on. We decided on running a python backend because even though our backend devs weren't particularly experienced in it, they felt more comfortable with python than node.js. Due to this being on such a limited time schedule we prioritized development speed over functional efficiency and python ended up being a good choice.

Challenges we ran into

We wanted to incorporate LLM feedback directly into the webapp without any hassle for the end user. A result of this was the decision that the LLM should also be capable of edition our codeblocks, just like the user. We solved this by building a json-based traderbot schema, to which the blocks would be converted into so the LLM would have a better understanding of what is going on and how to contribute. It's no secret modern LLM's are knowledgable about risk to reward ratio and how to give solid financial advice which we wanted to utilize to the fullest. A challenge nevertheless remained to keep the strategy blocks the user compiles dynamical which was achieved with object-oriented programming with a strong focus on re-usability similar to a custom programming language. This implementation came with it's own challenges however, but it was a solid learning experience.

Accomplishments that we're proud of

We are really proud of the teamwork we managed to pull off during the hackathon. Splitting the workload effectively and fairly was crucial to allow a good development workflow. Besides we are also happy with the project itself and our end result. We are proud to have a working demo and something to show for our work, even if it is an early prototype, the amount of functionality and usability is something we are happy with. A big priority and goal for us was to maintain ease of use and provide a nice user experience, which we think we accomplished quite well.

What we learned

All teammembers ended up working with technology they were not particularly familiar with. For instance MongoDB was something no one had used before but we were inspired to give it a try and it turned out teaching us all a lot. We initially wanted to include it for the learning experience but having our own database ended up being quite crucial for the performance of our final app. We also learnt a lot about building a two way psuedocode converter, which ended up sharing a lot of similarities with a traditional interpreter - programming this logic was a completely new experience for us.

What's next for AthenaAutoTrader

Our webapp is still in it's early stages and to finalize it's functionality and behaviour we would need to talk with someone more knowledgable about finance than ourself. The app is built in a highly modular system so it would be trivial to add more components and triggers, which would be essential if deployed to a larger userbase. Especially more advanced users would likely want a lot more financial analysis functions built in. We intend to open source the project with the idea of providing a starting point for anyone looking to make this into a fully functional service. There are also concerns about trading tools being excessively easy to use and if it could cause harm to someone who are not aware of the inherent risks with stock trading. The AI analysis tool is supposed to be a helpful tool to provide feedback and hopefully be a last resort safety net if someone tries making a bot that would make very poor decisions, but the bot is also not connected to anything with real money for the time being.

Built With

Share this project:

Updates