20190626_162554 Šnipiškės: a triangular brick house
20190707_104929 Šnipiškės: three kinds of buildings
20180728_095012 Machine Learning hackathon sponsors introduce their data science app hosting platform
Another Machine Learning hackaton sponsor was Domino Data Lab, which calls itself a Data Science Platform. They gave us a short crash course of how to use their platform. As far as I recall, you could deploy applications on it kind of like on Heroku or similar platforms, but tailored for data science. The details escape me now. The purpose of them giving free access to the hackathon attendees (but only for a limited time, like a month -- after that they were going to shut our apps down, and it was up to us to migrate them to some kind of more permanent space before then) was so that we could deploy our applications on it for the demo. (As is customary, the hackathon culminated in demos of applications the attendees created.)
Trying to use it, I was left with an impression that their platform was not very mature yet; it was more like a beta product. I hope it works much better now that over half a year has passed.
Again, to anyone who might be reading this, please don't take this as an accurate review of Domino Data Lab product. I'm sure the product must have improved a lot since then, though I never tried using it again.
This ties in with my general view of using some kind of untried platform for deploying your product at a hackathon. It is generally a bad idea. Usually, right up until the last minute the attendees are making changes to get that one last piece of functionality working right, just so their demo would look less like Powerpoint-ware and more like a live, interactive software. Almost nobody has time to also think how to deploy it to some new server and work out all the kinks of the deployment process. And that's true even if the hosting platform is something you are familiar with, such as Heroku! And there is not a snowflake's chance is hell that you'll be able to smoothly deploy it -- without hours wasted and heads banged against desks -- on an unknown, untried platform.
Similarly, if you are not already proficient in continuous integration / continuous deployment processes, a hackathon is not the right time to try it for the first time. Sadly, I've been at a hackathon (a different one) where a sponsoring company's representative took an hour to do a presentation on 12 steps of continuous integration to the hackathon attendees that were frantically working on the finishing touches to their apps. Some politely listened with their eyes glazed over, others (including me) just put their heads down and continued to work on their apps. Sure, when the demo is an hour away, that's exactly the perfect time to learn about an arcane 12-step CI/CD process and try it out for the first time! What could go wrong!
People usually do demos from their laptops, and that's fine.
Again, this is not a criticism of Domino Data Lab, or of the hackathon organizers. There is a chance that some of the attendees kept the DDL platform in mind as something to try out in the future. Knowing that this platform exists might have proven useful to them someday.