Inspiration
To forecast prices of commodities by taking into account various macroeconomic factors as well as events that might affect the supply chain by leveraging agents.
What it does
It employs 4 agents, one does a traditional time series forecast, the other brings in macroeconomic factors and geopolitical and other event data and validates it with the forecast of 6 months. if it is ok then the forecast is justified. The third agent goes over the web and does a check of the revenues sales and other factors left out. The fourth agent would combine all these 3 and gives a price recommendation strategy
How we built it
We built it using smol agents, albeit due to lack of proper message handling capability I wrote it from scratch.
Challenges we ran into
Making the agents combine with each other and employing 3 LLM's in one project
Accomplishments that we're proud of
Employing 3 LLM's in a single project and make them communicate with each other without any abstraction
What we learned
Agentic AI
What's next for Pretium Nova
Will try to implement it with financial markets.
Log in or sign up for Devpost to join the conversation.