Harsh Gosula

Research Intern @Indian Institute of Technology, Patna

Patna, IN

1.Developed a high-performance video anomaly detection model that achieved a ROC-AUC of 92.64% on the UCF-Crime dataset, while sustaining real-time inference speeds—setting a new standard for balancing efficiency and accuracy in surveillance applications.2.Designed a lightweight spatio-temporal architecture under the guidance of Prof. Maheshkumar Kolekar (IIT Patna), capable of detecting both sudden and gradual anomalies across diverse temporal scales within strict computational budgets.3.Co-authored a research paper based on this work, currently under review at the International Conference on Image Processing (ICIP) 2025, showcasing key architectural innovations, extensive ablation studies, and deployment-focused evaluations.