Explain how you would optimize GPU utilization for a multi-tenant AI platform at Postman to ensure cost-efficiency without sacrificing throughput.
