I don't know the access to resources, sheer scale, research, and good pay would make it pretty enticing for me. But really it comes down to what you are looking for long term. Do you want to optimize, finances, lifestyle, work in a particular space, learn more and eventually start your own thing? A job at Google probably scores well along most of these dimensions.
Understood, we will work on finding that info and surfacing it on the site. As far as monitoring and errors go, the program is actually hosted by us in this case. So unless the program is faulty in some way any response issue is probably our fault.
Regardless Google and Microsoft have offered a narrow slice of what can potentially be done in this space. They are still not seeing that we really need is a marketplace that has the ability to let machine learning specialists surface their work to businesses for a profit.
I just changed that to take the public key, thanks!. Users get a private and public key, the private key will work in all areas but should never be exposed. The public key can be exposed and in the dashboard you can white list the allowed origins.
>You have docs for accessing it but I can't find any for me building an api endpoint. The docs also seem to suggest I use a client library called "kittn", which I'm pretty sure is wrong.
No you are absolutely right. The docs are not built out yet and you are seeing the template data. We will curate this process and provide documentation. But a quick overview is as follows.
It shouldn't be an issue - the API currently sits on Google's Cloud Platform which provides the infrastructure and tools to scale the service. IMO Snapchat's usage of the platform really speaks to the stability and capability of it.
> - What separates you from existing API Marketplaces (e.g. Algorithmia)
Macgyver differs from Mashape in that we host the program and actually run the program against the data (processing). This allows us to have more control over verifying the up-time as well as any changes to the program which might affect an end user's application (we do our own versioning so any new changes will be pushed as an incremental version) as to mitigate negative impact to clients already using the API.
We differ from Algorithmia in that we don't require you to learn our development tools or programming languages. We offer custom runtimes. We function by specifying -
1. How we will supply data to your program
2. The runtime you want us to use
3. How we expect the program to respond
But ultimately we don't care how your program does that or in what language it is programmed.
> - Do you have programmer APIs other than JS?
Not sure what you mean by this, the API takes and returns JSON but it can be used with any programming language.
> - From your blog I am assuming you use docker for running the algorithms? Will the docker images be made open too?
The programs are executed inside docker containers. We are still architecting the optimal workflow for our publishers and if that means making the macgyver docker image public we will do that but that is still TBD. We may just offer a sandbox environment for publishers to test in.
> - How much effort do you think it would take me to transform an existing algorithm that I have to be available in your system?
It's likely going to be easier than any other tool. We are going to work on providing more documentation and releasing the API docs very shortly - but the gist of the process is that you can move your algorithm program (or directory) into a folder in a standard linux box and from there you will just need to modify how you get your input data and how you format your output data, and specifying a runtime for macgyver to call on your main program file.
> how much does Macgyver cost? Mashape takes a % of what I charge users
We allow you to name your own price and that is how much you will receive from those who use your program. However the client facing price will be higher as we tack on our additional processing overhead and a fixed percentage fee.