Developed time series forecasting models on NHITS and TFT architectures, improved accuracy by 30% Finetuned TimeGPT and Chronos LLMs for feature reduction, decreasing computation time by 40%