Scaling AI at H&M Group with Errol Koolmeister - #503

Model Complexity

Balancing model complexity is crucial at H&M, where simplicity often trumps complexity for practical reasons. Errol Koolmeister explains that most production models are relatively simple, like LightGBM, which are effective and easy to integrate into existing infrastructure 1. The focus is on achieving a positive ROI and maintaining simplicity while also investing in research for future advancements.

We try to keep it as simple as possible, but we're not 100% there yet.

--- Errol Koolmeister

This approach helps in scaling AI efforts efficiently, even if it means not always using the latest technologies 2.

   

Cloud-First Strategy

H&M's cloud-first strategy has been pivotal in its AI journey, allowing for flexibility and rapid deployment. Errol Koolmeister notes that starting AI efforts on the cloud eliminated many traditional IT constraints, enabling agile development across 250 product teams 3. This strategy facilitates collaboration between domain experts and AI specialists, ensuring that AI capabilities align with business needs.

We are 100% cloud first, which makes life a lot easier.

--- Errol Koolmeister

Centralizing data science efforts initially helped incubate capabilities, avoiding the pitfalls of isolated data scientists in business units 4.