Empower Your Data Scientists with Streamlined Delivery of AI & ML Services

Without Torque, data scientists who need to access, modify, or update the AI models or Machine Learning services are often required to submit a ticket to an IT, DevOps, or other team that controls access.

This creates an operational bottleneck that can slow down critical operational workflows for data scientists.

To resolve this bottleneck, Torque integrates with NVIDIA AI Enterprise and other AI tools to create, deploy, and monitor custom environments supporting the user’s specific AI and ML services.

With the environment managed as code, Torque can distribute self-service access for data scientists to access the outputs of the AI service on demand. Meanwhile, the IT and DevOps teams responsible for maintaining the environment can set custom policies, automate routine actions, enforce role-based permissions on access to models, and receive notifications about unexpected issues with the environment.

Also, once an AI service is deployed via Torque, the platform enables data scientists to “publish” it so that other groups can access it. For example, the infrastructure engineers maintaining GPU clusters and databases supporting the AI model can access the live environment to test and validate that their infrastructure and data are working correctly, while the developer building an AI agent or application can integrate the live model into their build.

This combination of self-service experience, collaboration, and governance helps to accelerate velocity for teams building AI agents while also preventing security and cost risk.

Torque is agnostic to the types of services, but some examples that our customers have used include:

• GPT-2 Large
• DeepSeek-R1
• Llama-3.2-1b
• Phi-2
• Mistsral-7b-Instruct
• Weaviate
• Spark
• Falcon-40b

Torque defines the custom code to launch the service, provides access to the outputs of the service, and makes it all available to data scientists via an intuitive self-service experience.

Other teams, such as development teams building AI agents or applications, can also access these services when provisioning their unique workloads. This makes access to AI services a seamless part of the development and management of the AI solution.

Yes! Torque leverages your unique NVAIE resources, allows data scientists and others to provision and access the outputs, monitors the state of these resources continuously to identify configuration drift or other unexpected issues, and automatically scales GPU allocation as your AI services transition through various phases of the lifecycle.