DecepTech AI is a machine learning-based lie detection system that deals with nonverbal behaviors. Traditional methods of deception detection involve polygraphs in relying much on physiological signals, which can be intrusive and may not give very accurate answers since individual variability causes them to go wrong. DecepTech AI avoids the problem by identifying deceptive behaviors based on a non-invasive approach of examining facial expressions, eye movements, and gestures. These cues are integrated using machine learning algorithms; hence the system offers real-time feedback during live interactions, such as interviews. The methodology applies to EDA that discovers the hidden patterns in non-verbal behaviors while creating predictive models for the classification of deceptive cues with higher accuracy and reliability. Features such as micro-expressions, gaze direction, blink rate, and gesture frequency are analyzed to enhance detection efficiency. The LSTM model achieved an accuracy of 81.52%, highlighting the system’s effectiveness in identifying deceptive behavior. The aim of this project is to construct a robust and bias-free lie detection system, testable in many domains with high precision adaptability to real-world applicability.