Enhancing data science with AI - Workcloud Modeling Studio

Nicholas Wegman, Senior Director AI Scientist at Zebra Technologies.

The Workcloud Modeling Studio offering

“At its core, Workcloud Modelling Studio is a low code/no code, data science-oriented, machine learning platform to build end-to-end, scalable forecasting pipelines,” explains Wegman. “We are really focussing on specific use-cases, and concentrating on demand forecasting as well as constantly factoring in scalability.”

Designed to streamline the creation and deployment of AI models, this platform empowers users with a low-code/no-code interface, enabling a single-click model deployment. By significantly reducing the time and technical barriers traditionally associated with AI projects, Workcloud Modeling Studio accelerates the return on AI investments. Its user-friendly design democratises data science, extending its benefits to a broader audience, including those with minimal technical expertise. This approach not only speeds up the AI deployment process but also broadens the application of data science across various sectors, making complex technology accessible to a wider range of users. This approach means that data scientists can spend more time doing what they do best, data science, instead of being bogged down in the code and scale.

Continues Wegman: “So if you think about a forecast generation for a retailer, they could have thousands of stores with hundreds of thousands of SKUs, which means you are quickly dealing with millions, if not hundreds of millions of things that you need to forecast, usually on a weekly basis, and sometimes even daily.” This is obviously a time-consuming process, and something that Workcloud Modeling Studio automates.

“This means data scientists can spend more time tuning parameters, analysing outputs, thinking of new models to run, and not just the mechanics of getting it to work.

“You can get yourself a pipeline that works very quickly, and then spend your time experimenting, as opposing to spending that time getting the code to work in the first place.

How it works

The Workcloud Modeling Studio offers a suite of AI algorithms specifically tailored for the Retail and Consumer Products sectors, addressing a range of challenges such as forecasting, demand planning, assortment optimization, allocation, and stock replenishment. This platform features ready-to-use AI pipelines designed for rapid deployment into live environments within a matter of weeks. These pipelines have undergone extensive testing, implementation, and optimization across various clients in the Retail and Consumer Products industries, ensuring their efficacy and reliability.

From an architectural perspective, the Workcloud Modeling Studio utilises the antuit pi spark libraries as the building blocks for its code. “The low code/no code aspect interacts with all these libraries that we have built, which we then leverage alongside the three big Cloud computing platforms (Amazon, Google, and Microsoft) to run everything off of,” Wegman explains.

From an input point of view, the model allows for you to input a wide array of data for different outcomes. “Of course, your sales data, as well as things like promotions and pricing, are going to be your most valuable data sets but you can also bring in lots of other important data points. Things like local events, weather, competitor information, or anything external to help improve your models.” All this information can be input at a rapid pace when needed to get forecasts back efficiently that accounts for all sorts of demand drivers.

Enabling data scientists whilst tackling business problems

A key strength of Workcloud Modeling Studio is its ability to better-enable data scientists to do what they do best, whilst simultaneously addressing concerns on the business side of things.

For experimenting data scientists, this system allows for an end-to-end pipeline to be set up and trialled within days and weeks versus the traditional months of waiting. It also allows these data scientists to make changes on the fly and effectively maintain these forecasting pipelines.

“The spirit of experimentation is absolutely at the heart of this, business stakeholders and data scientists have all kinds of ideas about what could potentially improve forecasts, and a lot of things don’t end up working.

“So, you really want to be able to try these things much, much faster. As opposed to spending three months trialling something and then coming to the conclusion it’s not going to give you better results, you can get to these conclusions in just days or weeks,” enthuses Wegman.

On the business side of things, Workcloud Modeling Studio maintains its ability to address the key areas of concern such as strong forecasting, replenishment, assortment, pricing, allocation, segmentation, and more.