Cassava Adds NVIDIA Nemotron to Its AI Factory

Cassava Technologies is enhancing its AI Factory by adding NVIDIA Nemotron models, giving businesses improved AI access, agent security controls, and tools to reduce computing costs.

The models will be available through Cassava AI Multi-Model Exchange (CAIMEx), a platform that allows businesses to access different AI models through APIs without having to host and manage the models themselves. The expansion also introduces NVIDIA OpenShell for controlling AI-agent activity and a Token Optimiser designed to limit repeated processing.

Also Read: Lucidya Launches AI Agent Platform to Scale MENA Expansion 

Understanding NVIDIA Nemotron Integration

CAIMEx is Cassava’s platform for accessing and using AI models. Rather than setting up the infrastructure needed to run each model independently, businesses can connect to models hosted and managed through Cassava’s AI Factory using APIs.

Cassava is adding NVIDIA Nemotron models to this offering. Nemotron provides reasoning and tool-use capabilities, allowing developers to build AI agents that can work through tasks and interact with business software. Its open model weights, training data and development resources also give developers the option to customise models for specific applications.

In practical terms, a business could use the platform to access an AI model for document processing or customer support without having to build its own model-serving system.

Securing AI Agents with OpenShell

NVIDIA OpenShell is a security layer for running AI agents in controlled environments. It uses isolated execution spaces and rules that restrict what agents can access, including files, credentials and external networks.

This matters because AI agents can do more than answer questions. Depending on how they are configured, they can use software tools, access information and take actions on a user’s behalf. Without appropriate safeguards, those permissions can create risks for sensitive data and business systems.

Cassava is using OpenShell to bring these controls into its AI Factory for customers deploying agentic AI. The aim is to let businesses use agents for more complex workflows while setting boundaries on their access and activity.

Optimizing Costs with Token Management

The CAIMEx Token Optimiser is designed to reduce the computing needed when AI applications encounter information that has already been processed.

According to Cassava, the tool identifies previously processed information and reuses it where appropriate, rather than processing the same content again. It is intended to work with both Cassava-hosted and third-party AI models.

Why does that matter? AI systems process information in units called tokens. The amount of processing required can affect the computing resources and cost involved in running an application. Reusing previously processed information where possible can reduce unnecessary work as usage grows.

Cassava has not disclosed measured cost savings from the optimizer, so its actual impact will depend on how it performs across different AI workloads.

Ahmed El Beheiry, group chief technology and AI officer at Cassava Technologies, said the expansion is intended to give enterprises and service providers more choice in accessing AI while improving the economics of deploying it.

Why it Matters to African AI Startups

For African startups, managed access to AI models could reduce the infrastructure work involved in building applications. Open models offer a foundation that developers can customize, while security controls become increasingly important as AI agents gain access to business systems.

Tools designed to reduce unnecessary processing could also help teams manage operating costs as their applications grow.

Cassava’s expansion brings these capabilities together within its AI Factory. The next test will be whether businesses can use them to move AI projects into reliable production environments while keeping security risks and computing costs manageable.