Assessment center consisted of two parts: Face to face interview with two people: Standard interview questions. One curveball question at the end designed, not to get the right answer but to see the process of how you think. Technical excercise with two people: Given some data about a choice to make with the available data and decide which option to go for. Have a "role play" where you present your findings afterwards. The role play ends and you may be asked questions afterwards. Only one person asked questions. Also a chat with graduates. My feeling is they want people who already know machine learning and advanced statistics techniques. If you know something that they don't know or haven't heard of, they won't hire you so don't bother. They want people who already know they stuff - but they are happy to train you in programming skills so basically if you have maths, stats background you should be fine. Engineers don't bother unless you've done advanced stats like POD.