DBTB INT Alya Abbot r

[Music] my name is alia Abbott I my company is up work and i'm a senior data scientist [Music] it's been really fun to just talk to a lot of people who are doing some more things add to what I've been working on and thinking about some of questions so some of it is just ideas for things to try that I haven't tried as some of this just kind of getting a sense of how other people are thinking about the same problems really powerful things we can we can do with with data so one of the great things about data at up work is that we have really complete information about the topic is an online feelings new platform so we have really complete complete information about the whole hiring process on the platform from when a user first sees the job to when they you know apply for it and how the contract goes what happens at the end so just having that complete funnel is really powerful for being able to be smart about picking out what we want to model what we want to optimize for it how we want to put it all together I would say one piece is that it's really important to think careful about what two models not just how you're going to approach the model what how you're going to do it but what what you want to model in the first place what you want to present what it means and then kind of how that fits in with your goals as a business also fear on the on the business side it's great if you can set your system up where you have the data that you need in order to be able to like have these choices and have these different ways you can analyze things so for example you know it up work we the data scientist will talk to the product managers about kind of how to make sure that they we have the data that we need for our models to work well and so what how it can make that happen so it can be kind of a back-and-forth process [Music] have a bunch of thoughts I guess one is maybe don't be intimidated like there's not like a great magical thing that all those data scientists now and it's a very new field right like five years ago nobody really called themselves the data scientist so you know you can't you can jump into it you can't figure it out I think I was just chatting with somebody about this today in terms of kind of four new people what I've seen sometimes that they don't really have they struggle with in terms of not having great intuition yet for sort of for the data so for things like how much data to need to get your model to work or like what's what's really important about modeling is it you know the features or was it a approach or kind of just getting us so so just getting that sense of like how to work with data it is I think really important as you're getting started so that when you really think that's true that's really something that distinguishes people who have like done this before it's not so much you know Dino particular technology like when I you know these days most of what I'm a lot of what I'm spending my time at work is writing sequels time at work is writing sequel scripts I had never ridden the sequel script in my life before I before I join it's not a big deal you can pick that stuff up in terms of I guess the kind of ladder question the sort of transit not the beginning knots getting started but like becoming great data scientist it's always hard to answer because I don't mean oh I don't know I think when it when I'm kind of trying to figure out what to do I there's often a trade-off between doing something that's kind of the local optimization versus a global one so just not in terms of the model but in terms of what I'm doing second spend my time like really figuring out like how to construct the model for a particular question how to clean the data how to understand the features and those kind of one piece where I could be and that's like I think that tends to be like what is going to make the best model the shortest amount of time for the task i'm doing and then on the other hand i could like go and say I'm going to spend my time like learning about deep learning and I'm going to bring in those methods and probably usually it's not going to bring the best results in the short term but it might be good for my career in the long term so I think there's kind of a maybe some balance you need to strike there and kind of I've ended up mostly like focusing on doing the right thing in this place and time where I'm at more so and then kind of going outside of work and reading up about other things and try to learn more but yeah I guess there's some there's I think some trade-off between like being great data scientists in terms of doing the best job you can right now versus like kind of setting yourself up for for future success [Music] you [Music]