“We’re using largely out-of-the-box capabilities and as a consequence of that, we also want to continue educating and providing learning opportunities to our teams, to the market, to the different people that are part of the McDonald’s system,” McDonald’s data science and data analytics senior director Patrick Baginski said, speaking recently as part of the virtual Data + AI Summit 2021. “We want to enable more end-to-end automation and machine learning operations, in general … and we want continue to implement governance, and also cost control measures in order to make sure that what we’re doing from the business perspective continues to make sense.” The Golden Arches took a serious leap into ML when it introduced DataBricks to its data science and machine learning users at the start of last year. As part of the process, the company learned about the best way to bring data into the platform.   “The way we do it is we bring in all the data into an s3 bucket where data lake is enabled … which helps us do data versioning and also build scalable and performance feature engineering pipelines in the platform,” McDonald’s data science and data analytics global director Abhi Bhatt said. Bhatt explained that within less than nine months of introducing the platform, the company went from zero to “production scale of ML Ops”. “We’ve not only identified the tools, the technology, we’ve done the legal paperwork, which can always be a hassle, but also identified use cases, built the models, and deployed them,” he said. He added that during this period, McDonald’s data science and analytics team managed to deliver more than 15 use cases, and sent more than 30-plus models into production before deploying them across five countries in which McDonald’s operates. Multiple ML deployment frameworks have also been built for the data science and end users, which Bhatt said was to ensure McDonald’s has the appropriate security frameworks in place for building and deploying ML models. Additionally, about 130,000 Databricks units have been used and roughly 27,000 compute hours on a monthly basis have been consumed, Bhatt said. But as McDonald’s looks to further expand its use of ML this year, he anticipates the numbers will grow at least four to five times. “That will be when we are truly at scale and doing ML Ops at McDonald’s,” Bhatt said. For Baginski, future ML work at McDonald’s will include carrying out “very fine grain, SKU-level forecasting” for its restaurants, automating marketing and personalisation-related activities beyond what he referred to as “good machine learning for marketing”, such as unsupervised end-to-end interaction with customers through offers, and cross-channel performance measurement across markets and restaurants. 

Databricks rolls out data sharing, automated pipelines, data catalogAtlassian claims it’s a step closer to achieving ’nirvana’ with its dataMcDonald’s next move may be hard for customers to swallow (if they notice it)HPE launches ML Ops container software service