Kratos GPT
Trained a VOLOv8 object detection model to identify and interact with 95% of in-game objects as Kratos in God of War: Ragnarok, using over 10,000 frames of video for model training. Leveraged Cohere’s command language model and GPT-4’s API to interpret strategy and decision-making using sentiment analysis and named entity recognition. Employed model quantization, pruning, and parallel processing to reduce computational demand and optimize model inference time by 35%.