goldsborough

It's very complete. We have many kinds of builtin Modules (rnns, conv{1,2,3}d, batchnorm etc.), optimizers, a data loader, a serialization format for checkpointing and similar things you need for training. My hope is that any model you can train in Python you can now also train in C++.

Web stuff, Python and Go

I learnt to be very careful about anything that's about Google and not from Google :)

No, I mainly used OCW to go deeper into more advanced dynamic programming and hashing techniques. I was more involved in the Princeton courses and especially Cracking The Coding Interview.

Yes and no. Interviewing is definitely something you can become good at by practice (I can confirm). But I can tell you that studying algorithms and data structures for interviews (which I would not have done otherwise) has made me a much better engineer and I've taken a huge amount of knowledge from it that I use very often. So it's more like one of those exams that taught you a lot, even if you don't remember everything after a year :)

No it was more a joke between me and a colleague, because even though we got gourmet food three times a day he would just eat potatoes and eggs all the time

Well, they did mean it for real. They didn't directly give me feedback on how to improve it, other than "go home, man". It's a personal choice in the end :)

In that midpoint form I mention there's a box you can tick if you want to do "conversion" interviews at the end of your internship. Since I still have two more years of uni I didn't tick that box.