If machine learning is to be used as a mean to make software systems more intelligent, the codebases for those systems need to adhere to software engineering practices too. Because when it comes to using machine learning in software systems it’s just a part of the big machinery consisted of several other modules. Therefore, to be able to maintain a sane, consistent and team-friendly development environment, it’s important for a machine learning (and deep too) practitioner to the world through the eyes of a software engineer.
My latest FloydHub article aims to give a brief checklist which a machine learning practitioner can refer from time-to-time in his own projects. The article tries to combine the pieces of advice collected from practitioners like Andrej Karpathy, Josh Tobin and at the same time includes many personal experiences.
This article also starts the mini-series on DL model debugging, tips, tricks, hitch-hacking and more! 
Looking forward to having your feedback.