AI and Chatbots and Models, Oh My: Why we Acquired Opteamizer

AI and Chatbots and Models, Oh My: Why we Acquired Opteamizer

November 30, 2022, is a day to remember: ChatGPT was unleashed onto the world, and we haven't been the same since. In just two months it acquired 100 million users, the fastest adoption rate in history.

Since then, AI in general and Generative AI in particular have been all the rage. Everyone seems to be talking about the disruptive potential of this technology, and how it will reshape our lives. In many ways, this is already happening. It seems that individuals can do anything with GenAI these days, from creating expert-level texts to generating out-of-this-world images.

However, leveraging GenAI and unleashing its potential at scale in the enterprise environment is a completely different challenge, one that requires making the most of High-Performance Computing (HPC) and adjusting the enterprise infrastructure.

Solving this challenge is necessary to enable enterprises, most still at the experimentation phase, to fully adopt GenAI, and is why we have acquired Opteamizer.

Scaling AI in the Business World

Despite the unprecedented enthusiasm about Generative AI (an enthusiasm I fully share), scaling this exciting technology is trickier than it seems.

Indeed, ChatGPT and other Large Language Model-powered (LLM) chatbots are a piece of marvel, able to generate high-quality responses to virtually any question, but it is important to remember when you are asking ChatGPT a question, you are already interfacing with a well-trained model that is focused on responding to prompts. Complex, modern businesses are a different story.

There are multiple challenges in scaling GenAI in a business context, ranging from issues of data, processes, reliability of output and more. However, for our purposes I will focus on one particular aspect: Models.

Implementing AI in your organization means, in simple terms, to create or replace a business process or function with an algorithm, whether it is to reduce costs, create new revenue streams or just to augment the organization's product or service. Most organizations either use models they have developed on their own or implemented an external model. These models often need large datasets and meaningful computing power. Luckily for us, what enables the new GenAI revolution is the abundance of both.

However, buying computing power and feeding an algorithm data is not as simple as you may expect. As business leaders, we cannot simply ask our tech leaders to “get me one of those”. In fact, it is important to understand that this issue is not only a technological one, but a complex business challenge. Companies that will not be able to effectively scale AI in their business will find themselves trailing behind those who can.

As business leaders, we have the obligation to understand the complexities of scaling AI and the implications on our business.

The crux of the complexity in scaling AI is that enabling next-gen models to operate at scale and in relevant timeframes (at times, the requirement is real-time or near real-time) is dependent on the model's architecture and the ability to leverage GPUs. While many traditional algorithms were designed for CPUs, modern business applications often require the parallel processing power offered by GPUs for specific tasks. Scaling traditional algorithms on CPUs at enterprise level can be resource-intensive, financially undesirable and technically daunting.

Therefore, to effectively scale AI performance at the enterprise, organizations need to better leverage GPUs, which is dependent on being able to either develop a specialized model, optimize an existing one or use an existing service (or open source).

In other words, you would need to optimize for GPUs. One could say that you need to Opteamize ??

The GenAI market is incredibly active, at times to the point that it is hard to keep up with new developments. Much attention is geared towards the competition between tech giants on creating the next-gen infrastructure of HPC. On the other hand, the onset of LLMs and edge visual processing capabilities have spurred a flurry of innovation and solutions at the consumer edge of the spectrum, with much attention being dedicated to the new feature released or to the new application made possible. An app that uses GenAI to make toast, anyone?

Finally, there is also a growing competition between tech giants and some newcomers on developing new feature on a regular base. This growing competition is both exciting and of global significance.

The picture that emerges is of a crowded space that’s hard to disrupt. However, a closer inspection reveals a fundamental gap: the ability to adjust enterprise infrastructure to enable models to operate at scale.

The truth is that if you are an enterprise today wishing to leverage the age of GenAI and HPC, you probably need to reconfigure the infrastructure available to you to make the most of advanced models at scale and in real-time (or in near-real time), as well as to manage it in a way that is both cost effective and delivers results. This does not avail itself to enterprises in most of today’s available solutions.

Opteamizer: A Missing Link in the Chain

Working with our clients, we've realized that this gap is not only real, it is also very acute and hampers the adoption of AI and GenAI in business environments. That is why we've decided to partner with Opteamizer and welcome them into the fold.

Established in 2015, Opteamizer is a team of mavericks, led by Tomer Gal ?. Years before GenAI was commonplace and at the crosshairs of every CEO, Tomer and his team were building models designed for HPC, helping their clients make the most of available GPUs. By partnering with NVIDIA and mastering the arts of CUDA, Tomer and his team are simply the best at designing models that are built to scale and to extract every bit of value from these high performing processing units.

Opteamizer will join our AI practice and will serve clients globally and will leverage their expertise and capabilities to solve the gap between infrastructure and application and will help our clients to scale their models in their environments.

We are confident that Opteamizer's expertise and capabilities will synergize with our demonstrated leadership in designing, implementing and scaling successful AI strategies and capabilities. Their addition builds on our recently expanded alliance with NVIDIA and demonstrates our commitment to investing in capabilities that will help our clients unlock the power of GenAI and HPC.

Finally, it is worth mentioning that our acquisition of Opteamizer is further demonstration of our commitment to enabling our clients’ efforts to create robust and actionable data strategies, to modernize their infrastructure and complete their cloud and AI journeys.

Tomer and team - welcome home!

?