The Indian Summer Monsoon (ISM) affects over 1.4 billion people, contributing 75–80% of India’s annual precipitation. Traditional statistical forecasting shows limited success at seasonal lead times, prompting exploration of machine learning methodologies. This study investigates relationships between Arctic Sea ice dynamics and Indian monsoon precipitation using 43 years of climate data (1981–2024). We incorporate Arctic Sea ice thickness and concentration metrics with conventional oceanic predictors through systematic feature engineering. Our methodology compares regularized linear models with ensemble approaches using temporal split validation. The Random Forest model achieved the best performance (R2 = 0.1256, RMSE = 12.14 mm, correlation = 0.366), suggesting modest but meaningful Arctic-monsoon relationships. While explained variance remains limited, tree-based algorithms capture non-linear relationships that linear approaches miss entirely.