Google Cloud’s consulting arm has announced the availability of a new offering called Generative AI Ops, which aims to help enterprises that are struggling to take advantage of the latest development in artificial intelligence and move new workloads into production.

The company explained in a blog post today that moving AI workloads into production is no easy task, requiring extensive knowledge of concepts such as generative AI systems design, large language model architectures, prompt engineering, evaluation and so on.

The problem is that the generative AI ecosystem has expanded so rapidly that there isn’t enough expertise to go around. Generative AI has become the number one priority for thousands of enterprises, yet two years ago hardly anyone had ever heard of it. It was only with the arrival of OpenAI’s ChatGPT, which showcased the technology’s incredible potential, that everyone suddenly became interested in it.

With its Generative AI Ops offering, Google Cloud Consulting and its partners are providing a solution to this lack of expertise, providing companies with access to a team of experts who will assist them at every stage as they transition their generative AI prototypes into production workloads.

According to Google Cloud, Generative AI Ops provides not only the expertise, but also an optimized technology stack for building AI, together with an extensive set of services for developing a wide range of models.

Generative AI Ops is said to focus specifically on the key steps required to bring generative AI into production. These steps include prompt engineering, design and optimization, which are necessary to ensure AI models provide high-quality, accurate outputs.

According to Google, its assistance covers techniques such as ReAct, retrieval-augmented generation or RAG, and “chain of thought.” The company explained that different AI models typically require very different prompting structures, so it’s important to employ the right techniques for each model.

In addition, Google will also lend a hand in performance and system evaluation. It explained that AI models must be continuously evaluated to improve performance and ensure accuracy. To that end, Google Cloud Consulting aims to help customers build an appropriate evaluation framework for each generative AI app.

The next step is model optimization and continuous tuning, which pertains to the ongoing work to improve generative AI models once they’re up and running in production. Google said its experts can help customers to optimize their AI system architectures, model selection processes, reduce latency and costs, and more.

Monitoring and observability is another essential capability Google is offering. The companies said it will work with customers to create the observability tools required to constantly monitor the performance of generative AI models to safeguard against inaccuracies and “hallucinations,” which refers to the way some models generate false outputs.

Finally, Google’s experts will also support business integration and testing to ensure that customers’ generative AI models integrate with their business processes in the way they intend. Tasks here include setting up cloud environments to host AI models, designing application programming interfaces to manage model interactions, and load testing to evaluate model performance.

Google said the Generative AI Ops program will be complemented by various training courses, hands-on labs, bootcamps and more within the Google Cloud Skills Boost Platform, so companies can train their own employees to get to grips with generative AI.

Image: SiliconANGLE/Microsoft Designer