Amazon SageMaker AI Studio adds no-code UI for inference recommendations

Amazon SageMaker AI has introduced a low-code/no-code UI for generative AI inference recommendations within SageMaker Studio. Previously, teams had to rely on the programmatic API, which assumed knowledge of parameters and raw benchmark output. The new UI removes that barrier by guiding users through preset use-case profiles—Interact (chat workloads), Generate (longer outputs), Summarize (high input-to-output ratios)—or a Custom profile for their own dataset and concurrency settings. Users then choose an optimization goal: minimize latency (for interactive apps), maximize throughput (batch/high-volume), or minimize cost. The UI also supports models from SageMaker JumpStart, S3, Model Registry, or existing SageMaker models.

Once configured, the system runs benchmark tests and displays visual comparisons of performance metrics across recommended instance types and serving configurations. Users can deploy their chosen configuration to a production endpoint with a single click, without writing any code. The entire process reduces the typical iteration cycle from weeks to minutes for common workloads and a few hours for custom ones. Advanced users retain access to the underlying APIs for fine-grained tuning. This feature is available now in SageMaker AI Studio under Jobs > Inference optimization, requiring only an AWS account with a SageMaker Studio domain and appropriate IAM permissions.