Lightning fast data analysis with pure numpy

We were able to speed up our analytics backend by a factor of 10,000 thanks to switching from pure Python to pandas to low-level numpy. Each transition compared to learning to walk again but was totally worth the effort. If you are interested in how pandas works internally, why it is sometimes 10-100x slower than plain numpy, and how to mitigate the implicit bottlenecks – this workshop should provide great value. We’ll study how typical data-fu operations such as groupby can be rewritten in plain numpy and whether you should or shouldn’t bother. Finally, we’ll learn how to use the next-gen Python profiler py-spy.