Mental disorder depression is a widespread mental health disorder that affects millions worldwide but most go undiagnosed for the limitation of the traditional screening methods. In this paper, we propose a NLP based chatbot and FER hybrid that provides more accurate and objective depression detection. BERT and Bi-LSTM are used to analyze conversational text in the chatbot for the analysis of sentiment scores, linguistic style and psychological indicators. At the same time, the facial emotion analysis module uses CNNs and Vision Transformers to recognize the depressive cues such as reduced eye contact, sadness, and lack of facial expressions. The proposed model uses hybrid ensemble learning model of XGBoost, SVM and deep learning based classifiers to fuse multi modal depression indicators to enhance the predictive accuracy and facilitate with robust decision making. The model is trained and evaluated on benchmark dataset and it gives significant improvement over traditional machine learning models. The integration of the textual and facial data shows slightly better depression detection accuracy which subsequently leads to reduced false negatives and false positives compared to the standalone methods. This AI powered methodology could assist mental health professionals with automated and real time screening tool for faster early diagnosis and intervention.