Walk me through the tradeoffs of synchronous versus asynchronous API designs when dealing with long-running ML inference tasks at Alloy Partners.
