Stock AI

With just a name of the stock, Buy or sell option and risk strategy, The platform provides you graph analysis(short, medium and long term), News analysis , sentiment and fundamentals analysis and lastly using all the information processed, an advisory strategy for the future. The platform fetches the real time stock data and analyses short, medium and long term graph data, and performs a visual analysis on it using Gemini Pro Vision model. We have GPT 3.5 32k model to analyze the news fetched from news API in the context of the stock. The graph analysis, news analysis, Fundamentals, user's strategy are all provided to a fine tuned LLM to provide recommendations on strategies for the user. With the evolution we plan to use an RAG with time series, where each time stamp with 5 days, 1 month and a year data is vectorized with the current value equal to zero and the rest time stamps value normalized. this helps the model to predict future time stamp by taking considerations into previous similar chart movements of k examples. This helps us to track the chart patterns without explicitly defining it. In summary we used RAG to semantically search k similar Graph patterns of a similar time frame and averaging the k graphs to predict the future chart movements. Feautures : 1. The models could be easily swapped. From GPT to Gemini in 5 mins 2. With fine tuning each agent the performance of the overall platform would rise exponentially. 3. Vectors created for the time series could be more advanced by including the Volume dimension as well as the news sentiment dimension