Use of AI techniques to solve real life problems have been growing rapidly. Number of technologies to handle these workloads has also seen an exponential growth both in terms of variety of hardware and software available. This growth in use cases and available technology options bring complexity to system designs targeted towards AI workloads. When designing the on-prem IT infrastructure to run AI workloads, it is critical to understand the impact of applications on the compute, storage and networking subsystems and utilize the right technologies for the target workload. This talk will describe how we approach system design optimized for AI infrastructure at Dell. The audience will get an understanding of how hardware, orchestration software, MLOPs tools and applications come together to form an integrated system with specific examples of projects from our CTIO and AI Innovation labs.