5 Critical Steps to ensure Cloud-Native Machine Learning Platform Success
By Bradley Clerkin, BreakFree Solutions CTO
Okay, you picked a cloud-native approach to data and ML platforms, now what? Firstly, congratulations on picking cloud-native. Your company and team will significantly benefit from this approach. If you’re wondering why I am congratulating you, check out this previous blog post about factory versus cloud-native. Assuming you are asking us where to get started if you subscribe to a cloud-native approach to data and ML platforms, here is what we recommend:
Find a business stakeholder who will engage with you on a consistent basis and who has a strong desire to get more from data. What insight or question regarding a single dataset or domain would be most valuable? Get them to give you just one. Make sure the question they want answered is really valuable and impactful.
Start a cross-functional team that can answer that question. If you’ve got a platform team already, great! Engage with them to give you XF team support in their journey with cloud platform systems. If you don’t, then we’d recommend adding some platform SMEs on your cross-functional team for now.
Now, get the team access to the cloud and the source data. Then, help them focus relentlessly on getting the answer to the first question in a minimally viable approach. The answer to this should come in under a month. Give them that deadline and don’t let it slip.
Once you’ve answered that question, determine the next three questions about that specific source data system. Your answers should have similar impacts and value. Give your team a similar deadline.
After clearly demonstrating the value, now is the time to take your efforts beyond MVP, advocate for more mature functionality, and focus on delivering self-service for the source data system. How can we get our users to ask their own questions? This may take a little bit longer, but its key is scalability.
Platform capabilities will enable quality and self-service. Next, expand the number of teams including establishing a dedicated platform that can evolve the platform for broader, repetitive consumption. This should be enough to get you headed in the right direction, for at least the first six months. I will be posting another follow up in my blog, but for the meantime, feel free to reach out at firstname.lastname@example.org.