Unlocking Enterprise AI Innovation: A Low-Code Approach with Mendix & NVIDIA NIMs™

AI adoption is skyrocketing, but most enterprise AI projects never make it past the pilot phase. The reason? Integration roadblocks, infrastructure complexity, and talent shortages. But what if you could develop and deploy AI without bottlenecks, without heavy coding, and without rebuilding your entire infrastructure?

In this article, we will demonstrate how the Mendix low-code platform can be used to create a simple AI chatbot that communicates with a locally hosted NVIDIA NIM.

Understanding the AI lifecycle

To implement AI, you need to begin by identifying a use case. Understanding how the chosen use case will add value to your organisation is important. An AI application needs three core components: data, an AI model, and compute. An organisation’s business, IT operations, and IT development processes also need to be considered.

Data is by far the most important consideration when beginning an AI journey. It is also worth remembering that an AI project can only deliver a business outcome when used in production, so identifying the problems that deliver the greatest impact to the business and have the least impact on your business, IT operations, and IT development processes is crucial.

What is Generative AI?

Generative AI, or GenAI, is a type of AI that can create new content, such as text, images, audio, and video, based on patterns learned from existing data.

In this post, we will explore large language models (LLMs), like Open AI GPT. An LLM is a model trained on vast amounts of data to generate responses that seem human. Open AI ChatGPT is an example of an AI chatbot application that uses the Open AI GPT family of LLMs.

The Generative AI Stack

When thinking about implementing generative AI, it helps to break it into layers:

Applications: 

Applications use LLMs and other foundation models. They are easy to use but offer limited control and flexibility and are often bound to a particular organisation's LLM models.  Examples include:

• ChatGPT

• Claude

• Copilot

Tools: 

Tools are used to build applications using LLMs and other foundation models. They are more flexible, allowing you to create custom applications. Examples include:

• NVIDIA AI Enterprise

• Mendix

• OpenAI API

  
  

Infrastructure: 

Used for training and inference of LLMs and other foundation models. AI infrastructure gives you complete control over where and how you run your models, which models you use, and how you use your data. Examples include:

• NVIDIA

• Dell

• Red Hat

• Azure

There are many ways to deploy an AI model, including cloud, on-premise, and hybrid. Each has different considerations depending on data privacy, cost, and latency.

What are NVIDIA NIMs™?

NVIDIA NIMs (NVIDIA Inference Microservices) are a way of deploying models as APIs that follow the de facto industry standard interface first developed by OpenAI. They allow an organisation to host their own LLM instance.

NIMs are containerised microservices that wrap around pre-trained models, making them easier to deploy and use. They can be run on any supported NVIDIA hardware in the cloud or on-premises.

This allows you to host your own inference services without relying on external APIs, giving you full control over data privacy, latency, and cost.

NIMs are part of NVIDIA's AI Enterprise platform and support more than 170 validated models, with more being added regularly.

Demo: Running an LLM with a NIM

In the video below, Simon Cottrell, Director of Technology at ebb3, demonstrates how to host an NVIDIA NIM locally with Docker.

Mendix Low-Code Platform

Many off-the-shelf AI applications only deliver around 70% of the functionality needed – you can buy or build. Customising off-the-shelf applications can be costly and time-consuming. However, many applications now have published APIs, allowing customisation and integration with other applications.

This means organisations can now buy and build. For many organisations, it is their customised applications that deliver competitive value.

Mendix is a low-code platform that helps teams deliver software faster and easier. It offers tools for building applications while allowing access to code where needed. For organisations looking to build AI applications, Mendix offers a visual development environment that enables easy customisation and extension of off-the-shelf applications. Registering for an account on their website allows you to start developing and prototyping applications for free.

Demo: Connecting Mendix to NVIDIA NIMs

In the second demo, Simon shows how easy it is to connect Mendix to a locally deployed NVIDIA NIM model.

The combination of Nvidia NIMs and Mendix gives teams a practical way to explore and deploy AI use cases. With NIMs, you gain flexibility and control over your models. With Mendix, you can rapidly build and evolve the applications that make those models useful. Together, they provide a streamlined path from idea to production

Infrastructure Considerations for Deploying Generative AI

Successfully deploying Generative AI applications requires a well-architected infrastructure, designed to meet the needs of your model. Here are some key considerations that need to be addressed before implementation:

• How many concurrent sessions need to be supported?

• How does the application scale as usage increases?

• What are the latency requirements for delivering responses?

• What GPUs should be deployed to meet performance needs?

• How many NIMs do I need to run?

• What servers are most appropriate for the workload?

• Can I use the cloud?

• How should I manage Kubernetes to ensure reliability and scalability?

These are all critical questions, and a successful AI deployment means getting them right. At ebb3, we have deep expertise in this space and work directly with our customers to design, deploy, and optimise infrastructure for AI workloads.  

If you’re planning to deploy generative AI in your organisation, get in touch with ebb3 to ensure your infrastructure is ready to support it.

Contact Us

To dive deeper into the topics covered in this article, including an exclusive Q&A session with NVIDIA, you can watch our webinar ‘Unlocking Enterprise Innovation: A Low-Code Approach with Mendix & NVIDIA NIMs below.

This is the first in a series of webinars we’re hosting in partnership with Mendix, so stay tuned for more sessions coming soon.