How to guarantee your machine learning model will fail on first contact with the real world.


Go to NumFOCUS academy page.

Recently I had my PhD thesis rejected. As a failure, I am uniquely positioned to recognize failure. While I am unwaveringly enthusiastic about machine learning, I aim to share my insights into failed machine learning modelling from real-world examples in science and industry. This talk is for you if you have an introductory understanding of machine learning and would like to avoid common pitfalls.


Jesper Dramsch

Jesper is a recovering geophysicist with a knack for coding, machine learning, and data science. During their PhD, Jesper focused on deep learning applied to geophysical data. During this time they finished 10 peer-reviewed papers in workshops, conferences, and journals. Being ever so distractable they also reaching top 81 worldwide in the Kaggle notebooks division and worked as a consultant teaching Python and machine learning in a few places. Recently they finished a book chapter, which is under review, titled “70 Years of Machine Learning in Geoscience in Review”. They were invited to rewrite the O’Reilly teaching material for Tensorflow 2 and have just published a data science with Python video course.