Inspiration
To get real time feedback on a stock trading decision you're about to execute or are considering.
What it does
Analyzes fundamental financial metrics and technical analysis for any publicly traded stock in real time. Additionally economic analysis.
Enter Ticker, Click Financial Or Technical Analysis, wait for analysis and sentiment (Sell, Hold. Buy).
How we built it
I built this in under 50 hours, using Replit and open ai api. It currently uses GPT-3.5 Turbo because it is much faster and cheaper than GPT-4. Additionally, I did not notice much difference between GPT-4 and GPT 3.5 Turbo in the analysis. GPT 3.5 Turbo seems to be a great LLM in itself for financial analysis on equities
Challenges we ran into
The challenges I ran into were mostly around the hard open ai input ceilings. Inputs and outputs are limited to 4k tokens. Financial data can get much longer than 4k tokens. This is great for short term analysis but if we were analyzing a longer time frame (1 year price movement of a stock), GPT-3.5 Turbo or GPT-4 would struggle with that. Even in their 32k token LLMs. Anthropic 100k prompt might be ideal
Accomplishments that we're proud of
The app is live and functioning and anyone can now search any publicly listed stock ticker and get real time financial/ technical analysis. We can also analyze the macro (economy) in real time! We store previously searched equities for a few hours at a time to shave off latency. Outputs can be as fast as 1 second or as slow as 30 seconds.
What we learned
I learned that LLMs are actually incredible at analyzing fundamentals about equities. GPT is familiar with all of the various terminology that an individual needs to be privy to when trading. It makes perfect sense why Bloomberg created BloombergGPT, this is a working man's version of that.
What's next for Paisa.ai
I will continue to harden the results, store them, and keep track of how we perform over time. I will incorporate a user base and implement a saas business model with different features for free/paid users.
Log in or sign up for Devpost to join the conversation.