We are not actually built on MTurk. The tagline is there to allow people to quickly understand what type of tasks we can complete.
We bring on our own workers that have been vetted and tested by us - and are also open to working with workforce providers. In terms of scaling, we are not concerned about the supply (workers). There are a very large number of workers available - and quality can be moderated with the right onboarding and testing processes. The more difficult side of the equation is the demand side. By building a simple RESTful API - we hope to be able to generate consistent demand.
We can handle both .mp3 and .mp4. The current pricing we have listed is the base price per api call for simple image moderation. Longer/more complex tasks will of course cost more. Feel free to fill out the form on the website and someone from our team will reach out to you with more details. Thanks!
Interesting idea. Sounds like you may hit scaling issues rapidly if this gains modest popularity though, since you're tied directly to a worker pool. I see there's about 300k HITs active on MTurk at any time, so even the startup pack at 1000 requests/mo would be a significant fraction of all MTurk activity.
Would you be expanding to other crowdsourcing providers as you grow?
Edit: Wasn't paying enough attention to the math. I guess without data like HITs resolved / second, it'd be hard to estimate just how many requests could be feasibly absorbed by the workforce. The rate limits look sane enough.
My first impression is that this is built on MTurk, but actually it's not. Then I don't understand what "Stripe-Like" is. Probably need to refine the headline.
I'm building a microwork platform but so far the main focus has been on the worker side and not on the submission of jobs side. Are you set on just using MTurk as your backend for workers or would you be willing to work with other providers?
If you're willing to work with other providers, what kinds of requirements would you have to ensure that the workers were high-enough quality for your needs?
I love this, simply genius. I think it'll make sense to focus on a few select tasks and scale those with great partners, instead of spreading too wide at first.
The pricing doesn't make all that much sense to me. Why not a simple per-request pricing model with a monthly minimum (after a # of free trials)? Or the ability to simply purchase larger blocks for discount pricing?
pkfrank - thanks for the comment. we are still working through the pricing model. The current model is on a per-request basis but we are also exploring a monthly fee to keep the quality of the companies that request tasks high. What kind of tasks do you think we should focus on?
Accuracy/Confidence is calculated based on the number of workers who have submitted an answer and the accuracy score that each one of those workers has.
In order to increase accuracy %, we simply push the task out to more workers. i.e. the exact same image moderated by 3 high-quality workers may have a 90% confidence level while pushing it out to 5 high-quality workers may increase that confidence to 95%.
We bring on our own workers that have been vetted and tested by us - and are also open to working with workforce providers. In terms of scaling, we are not concerned about the supply (workers). There are a very large number of workers available - and quality can be moderated with the right onboarding and testing processes. The more difficult side of the equation is the demand side. By building a simple RESTful API - we hope to be able to generate consistent demand.