Time series forecasting plays a pivotal role across diverse domains, often necessitating intricate domain knowledge and feature engineering efforts, which can be time-intensive and demanding. Deep learning has emerged as a promising alternative, offering data-driven methodologies to efficiently capture temporal dynamics. This talk introduces a new class of Transformer-based models tailored for time series forecasting. Leveraging attention mechanisms and integrating insights from classical time series methods, these models exhibit enhanced capability in learning complex patterns and dynamics. They demonstrate versatility, effectively handling both univariate and multivariate time series data. Empirical evaluations highlight significant improvements over traditional benchmarks, showcasing the practical utility of these advanced models in diverse forecasting tasks.