KalkiNi Tech

We are developing a violence detection system along with 4 more Ai models slated for deployment in a university context. Our current model utilizes the ConvLSTM approach, achieving an accuracy of 85%-88% on the benchmark dataset The current model is inefficient in achieving the goals in real time application and we are trying to use vision transformer as our approach to build model from scratch and make it an industrial based robust model ready to get deployed on real time. We seek professionals who have worked on vision transformers and can build the required models from scratch to achieve a good accuracy plus low latency by optimizing it and welcome the experienced people to work with us and build a safer future.