• Worked on key moment detection in video Livestreams (cricket matches) by leveraging inputsfrom video, chat, user, and audio modalities• Developed a novel fusion model involving multiple modalities and compared its results againstusing just the individual modes• Compared various feature extraction models like GoogLeNet (pretrained as well as retrained),Histogram of Oriented Gradients and Color histograms and improved accuracy of video mode• Detected lags between multiple modalities and incorporated lag removal in fusion model• Built a live demo to showcase the efficacy of our model.• Patent for work done in project filed with Adobe Internal submission