I support heads of Enterprise I&O and cover the general topic of “AI compute optimization” across multiple AI infrastructure layers. At the very bottom is the Facility and Energy layer including power and cooling; above that is the Compute Hardware layer including GPUs, custom ASICs, High Bandwidth Memory (HBM), high-speed interconnects, and CPUs; then is the System Software layer including compilers, libraries, memory management, etc., and the Service Orchestration layer including schedulers, inference engines, training frameworks. I do not generally cover topics above these layers regarding data and LLM models, devops, or AI services and applications.
I help clients to address questions like:
What are the key considerations to decide the best deployment environment for enterprise AI workloads?
How to optimize existing on-premises data centers to accommodate AI workloads?
What are the infrastructural planning and investment strategies to accomodate continuous AI stack optimizations and model improvement?