Cool! Like you, I wish people would make productive use of spare cycles.
Can I suggest you add a note comparing this to BOINC/gridcoin as well? I think your marketplace is a better idea, but because of the security implications dgacmu pointed out downthread, it shouldn’t be treated lightly.
Also, I really like your white paper (https://vectordash.com/primer) as it’s nicely formatted and has pretty icons. I think you should update the comparisons to the cloud providers though. Comparing a Pascal-series (1080 Ti) part to a Kepler part (K80) isn’t particularly fair, as NVIDIA charges about the same for that part as the later P100s and V100s. That is, people should be comparing what the best price/performance ratio is, and all three providers have better options. There’s a secondary question about whether the right comparison is to AWS Spot / Preemptible VMs / Low Priority VMs, but I’m not sure from a quick skim what the likely preemption rate will be with your setup.
Finally, let me echo dgacmu in saying it’s too bad that NVIDIA forces all providers to only provide the expensive parts. The best price/performance is certainly the consumer parts (that 1080 Ti compares favorably to a P100 at a fraction of the cost), but both the EULA and otherwise prevent it. You’re not running a direct service / providing things in a data center which is a cool workaround! But you should definitely be more clear about the risks of letting untrusted third-party code to execute against your GPU.
Again, great project!