The post is awesome. Having just "graduated" (dropped out of my PhD program) and also interviewed a bunch of really smart folks who are making the transition to

The post is awesome. Having just "graduated" (dropped out of my PhD program) and also interviewed a bunch of really smart folks who are making the transition to the "real world", I cannot emphasize the importance of the following for a good machine learning/engineering interview: 1. It doesn't matter if you have done research in a topic, do brush up on simple machine learning algorithms. The math after going through a PhD should be really easy for you. However, it is hard for an interviewer to gauge how good your math background is when you have done most of your quantitative work in a totally different field and you don't have enough of a common intersection. 2. Write code! If you are joining a small to middling start up, even if you will eventually be doing quantitative stuff, your peers will be people from engineering and they will evaluate you as a programmer. Brushing through something like the Algorithm Design Manual is going to be amazingly useful. Back in school, there was a linguistics professor who wrote most of his code in Scala, used version control and pivotal tracker. If someone who is working towards tenure can do it, there is no excuse for you. :-) 3. Try to work on tiny projects that involve some data analysis and some programming. If you are an R guy, do it in python and vice versa. Most importantly, think of the whole pipeline: taking the data (Infochimps/ AWS public datasets), cleaning it up, processing it using a random machine learning algorithm (pick any, its an easy and awesome learning opportunity), and then visualizing it. Repeat with cooler tools (E.g. d3.js for the visualization etc). The importance of this process is that it moves you away from the traditional anonymized datasets that you see in academia and helps you encounter real world data and gain intuition for that. This is incredibly helpful in the interview process because it allows you to get good ideas about other people's data and how to think about it.