• Worked on improving and optimizing the AI pipeline for video surveillance.• Conducted in-depth research to select an optimal up-to-date deep neural network for deploying on an edge device.• Employed optimization techniques such as model pruning, quantization, multiprocessing, frame filtering, and batch inference for optimizing resource allocation and accommodating scalability for multiple cameras.• Researched and evaluated open-source GenAI models, including DALL-E Mini, Stable Diffusion, CLIP, and various LLMs, to identify the most effective solution for retrieving videos from a large database using text prompts.