This study applies an artificial intelligence (AI)-based learning framework to analyze the return and volatility dynamics of four major foreign currencies the United States Dollar (USD), Euro (EUR), British Pound (GBP), and Japanese Yen (JPY)against the Indian Rupee (INR). Using daily reference rate data obtained from the Reserve Bank of India (RBI), the study covers the period from 28 February 2025 to 28 February 2026. Daily logarithmic returns are computed and modeled using a hybrid Autoregressive Moving Average–Generalized Autoregressive Conditional Heteroskedasticity (ARMA–GARCH) framework within a rolling window estimation approach. The results indicate that the USD/INR pair records the lowest annualized return (−0.4%) and volatility (6.8%), whereas JPY/INR exhibits the highest volatility (10.5%) and kurtosis (7.1), reflecting substantial tail risk. GBP/INR and EUR/INR demonstrate negative skewness and elevated kurtosis, indicating asymmetric return distributions and downside exposure. GARCH estimates confirm strong volatility persistence across all currency pairs, with α+β values ranging between 0.98 and 0.99. Furthermore, an LSTM–GARCH hybrid model improves volatility forecasting performance, reducing root mean squared error (RMSE) by an average of 21% compared to traditional GARCH models. The findings highlight the importance of adaptive, data-driven modeling frameworks for managing foreign exchange risk in India, particularly in periods of heightened volatility.