Otto: Friendly Machine Learning Assistant (github.com)

Now I have to distinguish between three things when I see 'Otto':

This project.

Hashicorp's dead deployment tool (https://www.ottoproject.io)

Otto Matic by Pangea (http://pangeasoft.net/otto/)

Is this the company that Uber bought then shuttered because of intellectual property theft?

Interesting project. Seems like a boilerplate to jump start various experiments.

Genuine question, because I'm interested in the applicability of a tool like this -- do professional ML scientists/developers have a need for visualizing their model during development, or would use such a tool? Do their models have parameters and need tuning such that visualization is a key problem?

Or is this kind of a nice to see thing for beginners, but more advanced/this-is-my-job ML professionals rarely are interacting with the data and model in this way? They're dumping tables, using numerical stats, not looking for visual patterns?

Perhaps a similar kind of question as "who uses drag-and-drop SAS analysis, rather than Python"? Or, "there are no casual DNN builders" -- anyone who is doing this kind of work will be doing it at the level that they're not needing visualization?

Clearly, SAS has enough visual-based casual users who can't program, so there's a market for it.

I am legit interested to know, thanks!

Interesting! How did you construct your network builder? I've had to use a combinatoric approach or constraint solver to validate layer compatibility but hit a wall trying to generalize it