
After working for weeks with Python’s and R’s data science stack I started to ask my self if there could be a common intermediate representation, similar to CUDA, that could be used by many languages. There should be something better than reimplementing and optimizing the same methods in each language. In addition to that, having a common runtime that could optimize the whole program instead of each function separately would be better. After a few days of researching and testing different projects I found Weld. To my surprise, one of the creators of Weld is Matei Zaharia, who also is the creator of Spark.