P. Rohit V. Acharya

Research Assistant @Texas Tech University

US

Developing a Multi-Modal Deepfake Detection System using Cross-Attention Transformers. Currently designing and building a deepfake detection system by combining visual, audio, and textual modalities using transformer-based cross-attention models (e.g, Multimodal Transformer, Hugging Face Transformers), achieving 95% detection accuracy. Processing and fusing heterogeneous data from datasets such as the DFDC, Fake News datasets, and VoxCeleb to enhance robustness against synthetic media.Applying Sequential Attention for feature selection on large-scale multi-modal data to improve precision and recall by 92%. Leveraging TensorFlow, PyTorch, and Scikit-learn for developing, training, and evaluating models with high F1-scores and real-time inference capabilities for cybersecurity and media forensics applications.Integrating temporal and contextual alignment techniques to synchronize audio-visual features, enhancing cross-modal consistency checks. Exploring adversarial training and contrastive loss methods to improve model generalizability under manipulated or low-quality content. Deploying scalable inference pipelines for real-time detection in social media and surveillance environments, ensuring reliability across diverse domains and languages.

Texas Tech University

Master of Science - MS, Computer Science

JNTUH College of Engineering Hyderabad

Bachelor of Technology - BTech, Computer Science & Engineering