A windows desktop app that submits images to Amazon MTurk for processing

Our CEO wants to have some desktop programs to demonstrate to customers what we do. Basically, we find cracks in things (walls, equipment, stuff like that) from photographs taken by other companies (e.g. drone companies). One of the ways that we do this is by submitting the job to the Amazon MTurk platform, where people in low-wage countries review the images. This program ("MTurk Mania") reads a config file that lists the AWS key, AWS secret key, the name of an S3 bucket, the HIT ID of the task we want performed (in the sandbox) and the HIT ID of the task we want performed (in the production environment). The user selects in the GUI whether they want to use the sandbox or the production environment. The program starts up, and displays one screen, which should look vaguely like the "MTurk Mania_1" image (attached). - At the top there is an area where you put in a price, a number of workers (usually around 3, sometimes 5 or 7), and whether this is "sandbox" or "live". - In the middle you have a list (initially empty) which have images and their current state - At the bottom there is a button "Go" which triggers the next step (and then disappears). The user drags images into the middle part of the list (or uses File -> Open). The program asynchronously uploads the image to S3 and shows a progress bar. (The Go button disables when an upload is in progress.) The image "MTurk Mania_2" is a rough idea, except that "ready for submission" should say "uploading". After the user hits "Go" the program creates MTurk hits (using the HIT ID defined in the config file) and updates the status of the image to say "waiting". It starts a one minute timer. Provide feedback on this so that you can see where it's up to (e.g. a progress bar that counts down to zero seconds). When the timer runs out, the program polls the MTurk API to see the status of all hits, and displays it accordingly. e.g. one image might have 1 response "no crack"; another image might have 2 responses, one saying "crack" the other saying "no crack". Once an image has the full count of assignments (e.g. if the pull-down at the top of the screen was for three workers), then it should also display a consensus. (e.g. if the majority said "no crack" then the consensus is "no crack"). The image "MTurk Mania_3" shows an indication of what this should look like. If there are any images that aren't ready, then start another timer and count down for another minute. (Also: after pressing go, you can drag any more images in, and the go button disappears). That's it! At this stage, we don't need to keep the results (we just display them on screen). There won't be very many images (maybe 10 if our CEO is feeling very enthusiastic) so there's no problem in keeping all these data structures in memory if it makes it easier. The primary goal is to have a nice-looking program that looks professional in a demo; it doesn't have to scale to support real-world usage. It just has to be useful in explaining to customers what sort of results we get from MTurk workers. I'm not overly concerned what language or technology you use to develop this, but I'll look very favourably on FreePascal/Lazarus, Qt (Python or C++ or Go), or F#. It has to run on Windows; if it happens to run on Mac or Linux that's a nice bonus. Include in your proposal what tech stack you will be using.