Thanks Zindi team for the great competition, I enjoyed working on it. The idea of partnering with AWS for computation was brilliant and I would hope we can have such initiatives in the future. Also, the instructions were quite clear (at least for me) to help connect.

I have a few suggestions regarding the use of an external computing server that I think could be helpful with any upcoming competitions that utilize the same idea and I encourage others to contribute with their own input as well.

- I would appreciate if we could have some information regarding the computing capabilities (GPU speed, memory... etc) mentioned in the competition description in order for us to have a realistic view of how much we can achieve in such an environment. I personally had too much expectations on how powerful the environment was leading me to use very large models that caused a lot of memory problems and the machine froze multiple times until I was able to get decent models that run without much trouble, my time was out and I was only able to submit results of a model after only one training epoch.

- It would be helpful to also include the contact of AWS (or any other service) tech support with the description so we can easily go to ask them about any trouble. Especially given the small time frame, it is very helpful to have answers asap.

- Maybe you could also add or link to some troubleshooting guide on what to do when common problems happened to help us save time with exploring different solutions to problems that occur frequently. It took me maybe 2-3 hours to fix the problem of frozen instance until I learned that I could just restart it from the console.

I wish this would be taken as it is meant to be, constructive criticism to make an already great idea even better.