The development of enterprise applications that leverage generative AI requires a comprehensive look and definition across several domains: AI, technology, and business. It is crucial to decouple decision-making in each area while maintaining a unified perspective on the solution.

The goal is to connect, customize, and generate intelligence for your business, while simultaneously integrating with your technology stack implemented with security, privacy, cost, and scalability in the architecture.

Achieving this end-to-end production architecture poses a significant challenge due to the need to merge these perspectives. It is a challenge for which we selected the NVIDIA AI Enterprise software platform for Globant’s Globant Enterprise AI platform.

NVIDIA AI Enterprise provides an “operating system” for AI, offering, among other features, several application frameworks and libraries for managing large language models (LLMs) from several providers (even your own model), and enabling scalable model deployment in production. Within this operating system, data can be prepared, models trained, and then deployed with optimized inference.

NVIDIA AI Enterprise 5.0 Reference Architecture Diagram. Credit: NVIDIA

On top of the operating system, the development of AI applications that meet diverse business requirements is necessary. Generative AI applications built with new NVIDIA NIM inference microservices included in NVIDIA AI Enterprise can be deployed anywhere – cloud providers, colocation data centers, on-prem private data centers, or certified OEM server manufacturers.

This is where Globant Enterprise AI comes into play, offering a business-centric approach to solution development. It connects existing foundation models powered by NIM from the NVIDIA API catalog with a layer of assistants designed to address specific use cases. This layer of assistants is accessible through various interfaces, facilitating the adoption of any foundation model available.

The API catalog will include community models as well as models from NVIDIA such as: Code Llama-70B, fuyu-8B, gemma-7B, llama-2-70B, mixtral-8x7b, mistral-7b, Nemotron-3 22B Persona, NeVA–22B, and more.

Globant Enterprise AI provides unique market abstractions for the creation of assistants by individuals close to the business. These assistants can interact with existing databases, documents, APIs, and analyze large volumes of information. Thus, Globant Enterprise AI is powered by the NVIDIA AI Enterprise platform and accelerates the creation of AI applications.

The architecture required for an enterprise-ready AI application is currently greatly underestimated. Globant Enterprise AI conceptualizes the following layers:

This infrastructure needs to be deployed across various environments, each layer must evolve, and non-functional aspects such as observability and security must be considered.

And when we talk about functional AI applications, we talk about solutions for all types of industries, including retail, finance, healthcare, and high-tech. This includes, for example, Globant’s MagnifAI testing platform that allows you to automate visual testing tasks using specific models from NVIDIA, or tasks such as generating test cases using the assistant layer given by Globant Enterprise AI.

About security and privacy concerns, Globant Enterprise AI enhances several of these aspects such as integrated access control models and integration with NVIDIA NeMo Guardrails, included in the NVIDIA AI Enterprise platform, and control input and output at the foundational model integration layers.

With this new integration, you can face the challenges of creating AI solutions by using NVIDIA AI Enterprise and Globant Enterprise AI in a cohesive way, taking into account all the details.

The journey to harnessing the full potential of enterprise AI is complex, requiring a thoughtful synthesis of AI, technology, and business strategy. Globant’s collaboration with NVIDIA represents a formidable approach to this challenge. NVIDIA AI Enterprise serves as a robust AI “operating system,” providing the necessary frameworks and libraries to manage and deploy LLMs effectively. On the other hand, Globant Enterprise AI builds upon this foundation with a business-oriented development platform that seamlessly integrates with NVIDIA NGC’s models, offering a suite of specialized assistants tailored to various business needs.

The integration of NVIDIA AI Enterprise and Globant Enterprise AI presents organizations with a comprehensive and nuanced toolkit to create end-to-end enterprise AI solutions. Leveraging this technology, businesses can confidently navigate the intricacies of AI implementation, better ensuring that their investment translates into tangible, strategic advantages in a rapidly evolving digital landscape.

You may also be interested in reading:

37 talks about software and Artificial Intelligence

Globant at GX30

10 Questions to Nicolás Jodal