Deepgram deepens Amazon SageMaker AI observability with Enhanced Metrics
Self-hosted speech AI carries an observability trade-off: the numbers that drive capacity planning and cost management stay locked inside the vendor container. Deepgram closes that gap on Amazon SageMaker AI with two capabilities that land billing, usage, and per-GPU metrics directly in your own Amazon CloudWatch account.
Deepgram is enhancing its Amazon SageMaker AI observability with two new features, aimed at providing billing, feature usage, and engine behavior insights. The first of these features, Deepgram Enhanced Metrics, allows for the direct publication of usage and billing metrics into a user's Amazon CloudWatch account. Notably, this feature does not require any additional agents, sidecars, or IAM permissions.
These metrics are the same consumed-unit values that drive AWS Marketplace metered billing, enabling a direct reconciliation of AWS bills against actual traffic down to the model and transport level.
The second feature, Prometheus and OpenTelemetry support, offers engine-level Prometheus metrics scraped directly from the Deepgram container, along with per-GPU and host metrics. These are collected through SageMaker AI's detailed observability and can be queried using PromQL from CloudWatch, Grafana, or any Prometheus-compatible tool.
The Deepgram Enhanced Metrics work within the constraints of SageMaker AI's deployment model, which isolates the container and prevents it from making outbound connections. This isolation makes it difficult to push vendor-level telemetry, but Deepgram's new features do not require the container to open a network path, and the data is still securely stored in the user's CloudWatch account.
The terminology, however, is slightly different - SageMaker AI has its own feature called enhanced metrics, which is distinct from and complementary to Deepgram Enhanced Metrics.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
