Open weight models, inference optimization, model serving, AI infrastructure, GPU performance, custom kernels, distributed inference, agentic AI, MLOps