NVIDIA NeMo microservices are now generally available. They help enterprise IT teams quickly create AI teammates that use "data flywheels" to improve employee productivity. These microservices offer a complete platform for developers to build smart AI systems and continuously improve them using business data, AI usage data, and user preferences.

What Is a Data Flywheel?

With a data flywheel, companies can add AI agents as digital teammates. These agents learn from user interactions and the data created during AI usage to improve over time — turning every interaction into valuable insight and better decisions.

Why Data Flywheels Matter for AI Agents?

AI agents need a steady flow of good data — from user interactions, databases, or real-world sources — to stay smart and reliable. Without it, their responses can become less accurate.

To keep improving, AI models need three types of data.

• Inference data (from AI usage)
• Business data (updated information from the company)

• User feedback (to know how well it’s working)

NeMo microservices help developers use all three.

How NeMo Microservices Help Developers?

NeMo microservices make it easier to build AI agents using tools that let you,

• Customize models
• Test and evaluate them
• Add safety and compliance features

Key NeMo Tools

• NeMo Customizer: Speeds up model fine-tuning by up to 1.8x.
• NeMo Evaluator: Makes testing models easy with just five API calls.
• NeMo Guardrails: Adds safety features with very little delay, only 0.5 seconds.

These services work with others like NeMo Retriever and NeMo Curator to build better AI agents, all using your company's own data flywheels. They're part of the NVIDIA AI Enterprise platform and can run on any supported system on the cloud or on-premises.

Why This Matters Now?

Businesses are creating multi-agent AI systems where many different AI teammates work together to handle complex tasks, helping employees do their jobs faster and better.

This shift shows that AI teammates could be worth trillions, helping with everything from fraud detection to shopping assistants and document reviews, all powered by smart data flywheels.

Real-World Results

Some major companies use NeMo microservices.

• AT&T: Improved AI agent accuracy by 40% using NeMo.
• BlackRock: Uses NeMo in its investment platform, Aladdin.
• Cisco: Built a coding assistant that is 10x faster and 40% more accurate.
• Nasdaq: Improved search speed and accuracy on its AI platform using NeMo Retriever.

Widespread Support

NeMo microservices work with many open models like Llama, Microsoft Phi, Google Gemma, Mistral, and more.

Meta has also connected NeMo microservices to its Llamastack platform, making it easier to build and run AI agents.

Many big software providers — like Cloudera, Datadog, and DataRobot — support NeMo, and developers can use it with popular tools like LangChain and LlamaIndex.

Build with Trusted Partners

Enterprises can run these AI agents on systems from Cisco, Dell, Lenovo, and others. Consulting firms like Accenture and Deloitte are also using NeMo to help businesses adopt AI agents.

You can download NeMo microservices from the NVIDIA NGC catalog, and they come with long-term support and security features as part of the NVIDIA AI Enterprise suite.