Why do companies want computer scientists for data science?
I understand that computer scientists have the ability to scrape and manipulate data but do they have any formal research design or statistical modeling experience? Wouldn't statistics or econometrics be a better fit for data science? You need to understand statistical theory in order to make accurate predictions based on sample data. Maybe my impression about companies wanting computer scientists for a data role is flawed. I would like to know your thoughts.
The chief statistic officer position doesn't exist yet so it all fits under the cio/cto.
"yet" you say?
Not in companies that still think all things technology related fall under a CIO
Which company and which language?
New company
Because model prototyping is only one stage of data science. Training on large dataset and productionizing models requires a lot more work.
What else? Would experience in python make up for the non-CS degree?
most data science folks don't have CS degrees, so that shouldn't be a huge concern. stats, math, applied math, physics phds all welcome as long as they could code.
Personal experience previous company director got a bunch of statisticians all phds
Nothing went to prod except pocs and were not able to get or look at the data unless given them as dataframes
The data was in hdfs.... may b exception but I saw 5 such data scientists so it depends not to be generalized
Companies don't understand what data science is. That's why. They lean towards infrastructure stuff versus finding real answers. It will pass.
Data science is code for CS + Statistics
Computer Science Research is all about math
Because data science has become so broad that people can try and fit any job description they want. A CS person plays a better role as they ( if having a good ML background ) are also the ones who knows how to love things to production, think about scalability , and other factors once prototype phase is completed
I think one should go and read the basics first . I will recommend to read the 21 rules by Martin zinkevich ( which has become like the Bible of ML models in prod from google ) The first rule which I have seen immense number of times in practice is to push a solution with minimal ML , basic mode to prod and get the benchmarks ( also tests the end to end pipeline works, specially when you have lot of dependent pipelines to get this job run)
Grad School I agree , research I agree , stats not too sure , there is always this divided view between what statisticians do and what machine learning / data scientists do ? So end of the day I feel what solves the problem and gets the job done . Now one can argue one of the best books in terms of fundamentals is still ESL and ISL , and stats have a lot of overlap with what our ML models do but things have come a long way today
They teach data science in CS programs. I basically "minored" in data science with the different courses I chose to take, plus you get hands-on experience rather than just theoretical. Computer Science is largely mathematics, remember, not just programming. Most of the business/math majors that I know who have gotten into Data Science roles have had to self-teach themselves to use the tools of the trade. CS bridges the gap for you.
CS isn't a coding boot camp