Multi-task deep learning pipeline for classifying emotions (angry, disgust, fear, happy, neutral, sad, pleasant surprise) and age group (young, old) from speech audio using the TESS dataset. Features include audio preprocessing (log-mel spectrograms), silence trimming, multi-head CNN model, Urdu translation support, and training with PyTorch