Deepgram integrates temporary AWS IAM into SageMaker for simplified speech models deploy

Deepgram has integrated AWS IAM's temporary delegation system into Amazon SageMaker AI, allowing developers to deploy speech models like Nova, Flux and Aura-2 with greater security and simplicity. The novelty includes a 14-day free trial period, eliminating the need for complex permanent permission settings. In practice, this reduces the setup time from hours to minutes, facilitating the adoption of voice AI in corporate applications.

What changes with the temporary delegation IAM to deploy speech models?

Before, to use Deepgram models in SageMaker, it was necessary to configure IAM roles with long-term permissions — a manual process and subject to security failures. With the temporary delegation, Deepgram's service takes on an AWS role for a limited period, following the principle of minimum privilege. This means that credentials expire automatically, reducing leak risks and simplifying access management. The improvement is aligned with best cloud security practices and was detailed in the official AWS blog(https://aws.amazon.com/blogs/machine-learning/deepgram-enhances-amazon-sagemaker-ai-support-with-aws-iam-temporary-delegation).

Speech models Deepgram available on SageMaker AI

Three Deepgram tip models can now be deployed via SageMaker AI with the new support:

New: real-time transcription model with low latency, ideal for call centers and live subtitling.

Flux: specializing in natural speech synthesis, used in virtual assistants and content narration.

Aura-2: Language model focused on voice interactions such as conversational chatbots.

All can be tested free for 14 days at no initial cost, directly from the SageMaker marketplace.

Why does the simplified deploy of speech models matter to Brazilian developers?

The simplification of deploy reduces technical and financial barriers for startups and companies in Brazil. With the temporary delegation, a dedicated DevOps team is no longer required only to configure AWS permissions — a solo developer can deploy and test voice models in minutes. The 14-day trial allows validating the quality of transcription or synthesis in Portuguese before any compromise. In addition, native integration with SageMaker AI facilitates horizontal scalability, something critical for applications that need to process large volumes of audio, such as service centers.

Companies that already invest in AI to optimize operations — such as Coca-Cola, which implemented Adobe system to manage brand in 200 markets — can benefit from robust voice solutions without restructuring all cloud architecture.

How to test speech models in SageMaker AI with temporary AMI?

The process is direct: access Amazon SageMaker AI, look for "Deepgram" in the template marketplace, select the desired model (New, Flux or Aura-2) and follow the configuration wizard. During the flow, an IAM role with temporary delegation will be requested — SageMaker himself offers a template for this. After validating, the model is available as an endpoint and can be consumed via SageMaker's standard API. The full documentation is on the AWS blog.

Frequently Asked Questions about Deepgram speech models in SageMaker

Does the 14-day free trial have any usage limitations?

No, during the period the use is completely functional, without restrictions on the volume of requests or duration of sessions — only the 14-day period is fixed.

Do I need an AWS account to use?

Yes, an active AWS account is required to access SageMaker AI and configure temporary delegation permissions.

Does the temporary IAM delegation really increase security?

Yes, because it eliminates permanent credentials, reducing the attack surface and ensuring that permissions expire automatically after use.