A full set of AI libraries for prediction from financial time series, built on specialized LLM and other machine learning models. That specialization is what allows us to go considerably deeper into the structure of financial data than any generalized model can.
Advanced mathematical techniques yield a full probability
distribution over outcomes
P(outputs | inputs)
for inputs and outputs of arbitrary dimension,
across both continuous and discrete variables. Particular care has been given to modeling
the correlation structure.
The resulting predictor then serves as the foundation for portfolio construction and trading algorithms.
Built with the Quantaiko AI Library Agent.
Below is an example for cryptocurrencies: the joint distribution of Bitcoin and Ethereum returns 5 minutes ahead, predicted from several past returns, in minutes in the figure below. For example, BTC_ret10 is the Bitcoin return between t−10 minutes and t.
Input state — average backward returns (%). Select a row to plot its distribution; drag the plot to orbit, scroll to zoom.