What would you have as some of the key skills for type 1? Would good knowledge of multivariate statistics be a critical skill? Asking because this is a subject that is close to my heart, and learning it out of my own interest, but cannot gauge how important it is to be a good data scientist of even a data analyst.
On the topic of hiring someone with both type 1 and 2 skills, why not hire both, given that they seem to be a mutually exclusive? And have type 2 pick up from what type 1 develops as hooks and use the hooks to productionize code?
This is a fascinating subject. One that I've thought about a lot. I started my career as a type 1 and transitioned into type 2 because i found type 1 work not scalable and subject to whims of management.
Type 2 work is much more scalable and permissionless.
As for your condidndrum, I'll be the devil's advocate and say this is a false dichotomy.
Type 1 is a thinking paradidm. People who can think clearly already intuitively grasp stuff like statistical significance, survivorship bias, confounding variables,simpsons paradox etc. I've worked in finance where traders grasp these things much better than any data scientist I've ever met because they put money where their mouth is. Just that they dont label this stuff. If someone is smart, they already know this. You can train them to label it and productionse it.
i guess my difficulty has been that most type 2 people i've interviewed (while in my previous job) have been extremely bad at type 1 stuff.
takes some amount of "business background" to have an intuition for type1 (hence traders, etc. get it easily). and a LOT of data scientists nowadays come from software engineering backgrounds.
it's not funny how quickly our hiring funnel used to narrow after a simple Bayes Theorem question in the first interview.
That's my experience as well. It needs exposure to the real world to get good at that stuff, but MLEs in startups have this, IMO. That could be a good segment to target.
I think a lot of smart people who wanted to work with data have abandoned data science (type 1) work for machine learning engineering(type 2s). I have seen a lot of people make this transition - rarely seen the other way.
So, in trying to start with type 1, there is also some adverse selection happening.
now that I've read Dhruv's comment here, I think I tend to agree wtih this. Now I'm wondering - as a startup wtih single digit people, can I afford to hire pure type1s, and wait for them to become productive as type2?
because this didn't work wtih me - i have a special ineptitude for "systems programming", and os have never become good at software engineering.
How do you evaluate a type I data scientist for type II requirements?
It is interesting because I also bucket myself in the type I role, but it is essential to have the knowledge and the ability to take up the responsibilities of a type II data scientist.
i really don't know - and since i'm type1 myself, i find it unintuitive to test for type2. i guess i just look at the kind of words people use. i simply ask "how good are you at production" and look at how they answer (not easy to bluff on this), etc.
What would you have as some of the key skills for type 1? Would good knowledge of multivariate statistics be a critical skill? Asking because this is a subject that is close to my heart, and learning it out of my own interest, but cannot gauge how important it is to be a good data scientist of even a data analyst.
On the topic of hiring someone with both type 1 and 2 skills, why not hire both, given that they seem to be a mutually exclusive? And have type 2 pick up from what type 1 develops as hooks and use the hooks to productionize code?
This is a fascinating subject. One that I've thought about a lot. I started my career as a type 1 and transitioned into type 2 because i found type 1 work not scalable and subject to whims of management.
Type 2 work is much more scalable and permissionless.
As for your condidndrum, I'll be the devil's advocate and say this is a false dichotomy.
Type 1 is a thinking paradidm. People who can think clearly already intuitively grasp stuff like statistical significance, survivorship bias, confounding variables,simpsons paradox etc. I've worked in finance where traders grasp these things much better than any data scientist I've ever met because they put money where their mouth is. Just that they dont label this stuff. If someone is smart, they already know this. You can train them to label it and productionse it.
I would hire type 2 people who have this sense.
i guess my difficulty has been that most type 2 people i've interviewed (while in my previous job) have been extremely bad at type 1 stuff.
takes some amount of "business background" to have an intuition for type1 (hence traders, etc. get it easily). and a LOT of data scientists nowadays come from software engineering backgrounds.
it's not funny how quickly our hiring funnel used to narrow after a simple Bayes Theorem question in the first interview.
That's my experience as well. It needs exposure to the real world to get good at that stuff, but MLEs in startups have this, IMO. That could be a good segment to target.
I think a lot of smart people who wanted to work with data have abandoned data science (type 1) work for machine learning engineering(type 2s). I have seen a lot of people make this transition - rarely seen the other way.
So, in trying to start with type 1, there is also some adverse selection happening.
I think if you put a type 1 in a production environment for long they'll become those pricy unicorns you talked about.
now that I've read Dhruv's comment here, I think I tend to agree wtih this. Now I'm wondering - as a startup wtih single digit people, can I afford to hire pure type1s, and wait for them to become productive as type2?
because this didn't work wtih me - i have a special ineptitude for "systems programming", and os have never become good at software engineering.
How do you evaluate a type I data scientist for type II requirements?
It is interesting because I also bucket myself in the type I role, but it is essential to have the knowledge and the ability to take up the responsibilities of a type II data scientist.
i really don't know - and since i'm type1 myself, i find it unintuitive to test for type2. i guess i just look at the kind of words people use. i simply ask "how good are you at production" and look at how they answer (not easy to bluff on this), etc.