DBTB INT Sanghamitra Deb r

my name is Sanghamitra Deb I work for Accenture technology labs as a data scientist in the machine learning natural language processing area well I'm still about to speak but in general I think the exciting parts about data by the bay and the entire conference was to kind of recognize the cutting edge spaces where text or NLP kind of work is happening and also a recognition of the fact that most people agree that models are not the most important part but rather getting the right data getting the right labels cleaning the data and that part of the work is way more important than the cool of machine learning part and that has been my realization well there is a lot of data it's kind of spread around so one of the jobs what makes it challenging is also to assimilate all the data and get like a wholesome insight out of it and it's exciting because it's everywhere like it's there in health care it's there in ad targeting it's there in search relevance and it's just exploding there's a lot of information there you've only teased out a tiny bit from it and it's exciting that there's so much to do well I'm going to speak about extracting medical attributes from data and the most interesting and challenging part about it is how such extractions can fuel personalized medicine so I'm extracting things like what are the side effects of drugs what are the diseases that it creates what new disease could a treat what age group are you are you targeting so all of this and I'm also looking at data that's already there so I'm not really collecting new data or rather working with public data and the thing that's come I'm going to come out of it is like a 200-pound man with some disease will not get the same medicine as like someone like me 100 pound person with the same disease but a completely different ethnicity or gender and other thing so it's it's an age where not just my work but all the talks that I'm hearing today where the different these are like the different blocks that are coming in do you feel personalized medicine yeah it's ironical that you asked me this question because so I spent a lot of last year giving talks on how to become a data scientist from an astrophysicist because I did astrophysics before getting into data science and my view on data science is it's a very generalized field it's and I may take like the opposite view of a lot of people and it has a lot of aspects are there is math there's statistics there's computer science there is a business side to it and then there is like the specific field you're looking at you may be looking at Pharma or you may be looking at market research or you may be looking at product and there's like the other part where the specific field you're looking at and it's very hard to be good at everything even as a company if you're looking for someone who's good at all of them it's really hard so it's important to pick a few but you have to be more good at more than one thing so you cannot just be a very good statistician and say I mean you could still say you're a data scientist but not be very good at it so maybe figuring out if you're new you know you're starting out from a different field figuring out where your strengths are for instance for me the strengths were quantitative so I chose the parts that were quantitative and I kind of honed on it and I had done a lot of talks in academia so I also chose communication as one of the things that I wanted to work on and like choose a few from these and work on them and find like the right fit on the other side you