NVIDIA NeMo accelerates enterprise AI agent deployment

NVIDIA’s NeMo microservices are now generally available to help businesses rapidly build, deploy and continually optimise AI agents that can work alongside employees to boost productivity and efficiency.

The new suite of tools lets developers create advanced agentic AI systems that learn and improve over time. By adopting a “data flywheel” approach, these AI teammates can use real-world user interactions, business data and AI inference results to continually enhance their performance.

Every interaction with the AI agent helps make it smarter and more effective, turning day-to-day usage into valuable training data that refines the agent’s capabilities and automates increasingly complex tasks.

Among the key features of the NeMo suite are NeMo Customizer, which accelerates the fine-tuning of large language models and delivers up to 1.8 times higher training throughput, and NeMo Evaluator, which simplifies the evaluation of AI models and workflows through streamlined API calls. NeMo Guardrails strengthens compliance and safety, offering up to 1.4 times better protection with minimal impact on latency.

These microservices can be combined with NeMo Retriever and Curator to build robust data flywheels, ensuring that AI agents remain accurate, efficient and up-to-date as they process new information.

Early successes

An early adopter is AT&T, which has reported up to a 40 percent improvement in accuracy for AI agents handling tasks such as personalised customer service, fraud prevention and network optimization.

Cisco’s Outshift team has experienced a 40 percent reduction in tool selection errors and up to 10 times faster response times with their coding assistant, while Nasdaq has achieved up to a 30 percent increase in search accuracy and response speed on its GenAI platform.

The rise of AI agents represents a trillion-dollar opportunity with applications ranging from automated fraud detection and shopping assistants to predictive maintenance and document review. The ability to build strong, continuously updated data flywheels is essential for turning enterprise data into actionable insights and scaling the impact of AI teammates across industries.