ML engineers can't write production level scalable code. They're always boasting about the accuracy of their solution. Some can't even tell the difference between a GET and POST request. AND ITS SO HARD to get them to admit they're wrong. 🙄

Because they'll be writing services which will do real time predictions on data.

If you're making ML engineers work at the level of HTTP requests, there's something wrong there. While I fully support having basic knowledge of engineering in general (of all kinds actually), this should be handled by a different kind of specialist.

@molaram you know ML/DL works and is deployed in production all over the place, right?

@RememberMe I was about to say this.

@molaram ML esoteric pseudo science? interesting choice of words.

@RememberMe yeah i agree, there needs to be a different kind of expert who can handle production level stuff and let ML people do algorithm development.

@rjcrystal how about...web backend developers? You let the ML folks export their model with an easy-to-use interface like model.classify({input data from request}) and you let the backend team treat it like a black box that just works (tm). The ML model shouldn't care about how the data is produced or transmitted, it should just expect a stream of data coming in and generate a stream of data going out.

@RememberMe makes a lot of sense to me. It'll make everyone's lives easier at work. 😁

Probably because that's not their job. Seems like you have the architectural flow backwards; here we consider it the responsibility of the engineering teams to deliver a platform that will consume ML models. The data science teams will have Data Engineers who serve as the glue between the data scientists and the systems that use their work. It's just basic human skills specialization.

And I mean, when you can get $2/hour resources from IBM india to write the GET and POST functions, why would they have $100/hr data scientists do that work? It's beneath them, and if someone asks a data scientist to do grunt level programming, don't be surprised when they quit and go to a better job.

@SortOfTested it could be difference in their expectations specifically from a structured and specializing perspective I guess.

Also there are different processes and philosophies in different organizations. Outsourcing can get the straightforward jobs done easier and cheaper even. But its also disconnected. I believe cross functional teams can bring greater value.

@molaram Nothing personal, but Teslas are better drivers than most humans with a driver's licence. Not 51% but at least 85. This thing works, better accept it and think about what skills you are supposed to expect from someone whose primary field of expertise is statistics.

@molaram lol nah man its cool. I just never heard anyone calling probability, statistics, linear algebra and calculus "pseudo-science". Ever seen how dogs, or most animals actually, react when you make a weird sound around them? damn near 99% of the population of engineers/scientists will look at you like that for saying that math is pseudo-science
:P

@molaram same :( i really dislike how people are using it for the most part to just sell shit and do advertising better, seems like a "beneath me concept" for it really. There are some cool applications where some people are using it for more than predictive analysis tho! it can be used to fine tune systems for a particular use and that shit really is innovating. But so far it really just seems like people are using it more for marketing stunts and that makes it sound shitty and superficial