We participated in a small evening seminar at Lift99 to listen Bolt’s Senior Data Scientist who shared some interesting thoughts about his job – stuff that they don’t teach you at school (for real 🙂 )
How did Bolt start with a data science department?
They pitched to their management with 3 people from the analysis team to form a separate department, which now has grown to 30 people.
How can you differentiate a data analyst from a data scientist?
In Bolt, the borders are a bit fuzzy, but in overall – the analyst deals with the information already obtained and how report it. The scientist works with questions & hypothesis while seeking ways how to prove them (e.g. how to improve the accuracy of the waiting time for the vehicle?).
A note to mention: You shouldn’t ask any finished tangible assets or exact probabilities for any research before the research has been done. Mainly because this is the scientist’s work to find out the probabilities and possibilities through his/her work.
Like in Spotify, roughly 95% of data science projects fail to prove set hypothesis’s but within the 5% of success, it’s a significant win.
On pictures:
1. (above) Machine Learning’s (ML)Â part in a “product” is actually really small, compared to data gathering, analysing, infrastructure creation etc. Although the expectation is the contrary (the green part of the bar).
2. (below) Research, where they mapped data scientists’ expectations and reality of working with ML. It showed an exceptionally big gap as people predicted that when working on a ML project you mostly work with optimising the model when actually it’s a really small part of the overall work. Most of the time goes to normalising and cleaning the data.
If you want to discuss any ML projects, do not hesitate to contact us by clicking the button below.
