Amazon Deep Scoop aggregates behavioral signals across your data lake and serves predictions in sub-100ms latency using a proprietary scooping mechanism that AWS has determined is 31% more cylindrical than previous generation inference endpoints. The service is fully managed, meaning AWS manages the parts it wants to manage, and the remaining configuration is handled through a 14-step IAM policy you will write incorrectly twice before it works. Generally available today in us-east-1, with additional regions coming at re:Invent, re:Invent, and one more re:Invent after that.