I would (and am) going into MLOps. Not just the general infrastructure/systems administration but how to do inference optimization, caching, quantization, memory pinning, vfio passthrough of gpus etc.