I did not work in the hottest topics in Machine Learning (and not in computer vision) as I cannot find rooms for innovation. I did publish several papers in 2nd-tier conferences and submit papers to journals. Near the end of my graduation, I started to find jobs in the industry. I have been shocked when I saw some employers will target authors who published in these big conferences even for non-research positions. (It sounds like discrimination.) I always wonder if my research profile will meet their requirements.
It seems like you are talking about publishing research work in book chapter rather than peer reviewed articles.
A book usually covers a selected topic in depth and in a coherent manner. This is particular useful for graduate students. Besides, some book authors also invite their friends to give comments. The impact of a book can be as rigorous and significant as journal articles.
By showing that almost no recruiter check the GitHub repos cannot tell if such contributions make a candidate stand out.
If someone has tangible contributions through GitHub, they probably write that down in the resume. The interviewers may check their claims either during interview or on GitHub before/after the interview.
The problem is that if a candidate performs moderately well in the interview (e.g. coding interview), while he/she has made many contributions through GitHub, should the interviewer recommend the candidate? My belief is that most interviewers will not take the risk or take the responsibility for a potentially wrong hire.
This is similar to the presumption of innocence. You can believe someone acts in good faith initially. Yet, if they don't, you should turn against them immediately to protect yourself.
How often will the clients find you again to do the follow-up work, e.g. new software feature requests, for the previous projects? If you refuse the requests, will this undermine the business relationship?
Nowadays, many books cover the elementary mathematics in machine learning. After I learnt these elementary topics, any good suggestions for computational learning theory?