Sign up  |  Log in

Dream job....turns out to be a dead industry

SamCryBaby wrote:

Either you know OP has a decent quant background or you are speaking out of your ass.Data Science ( such a stupid name ) and ML need extensive backgrounds in CS,Math and Statistical techniques let alone domain knowledge. By taking courses online you don’t become a date scientist.

I agree that by taking a course (or courses) you won’t instantly become a Data Scientist – but I disagree that you need an extensive background to learn the basics of Data Science and Machine Learning.  After all, in today’s world if one isn’t open to learning new skills and technologies then you can expect your career to go nowhere (and eventually you’ll be automated out of a job).  Of course you won’t instantly be at the cutting edge of Deep Learning & Neural Networks – but the libraries within Python or R can do so much heavy lifting “under the hood” to the extent that you can run a basic Machine Learning algorithm with literally just a few lines of code.  The danger there, of course, is that the user should have a good understanding of the mathematical / statistical underpinnings of the algorithm to make sure that the model is valid – so, I do agree that a mathematical / statistical background is necessary… but I’m not sure that it needs to be “extensive”.

Marathon_runner wrote:

SamCryBaby wrote:

Either you know OP has a decent quant background or you are speaking out of your ass.Data Science ( such a stupid name ) and ML need extensive backgrounds in CS,Math and Statistical techniques let alone domain knowledge. By taking courses online you don’t become a date scientist.

I agree that by taking a course (or courses) you won’t instantly become a Data Scientist – but I disagree that you need an extensive background to learn the basics of Data Science and Machine Learning.  After all, in today’s world if one isn’t open to learning new skills and technologies then you can expect your career to go nowhere (and eventually you’ll be automated out of a job).  Of course you won’t instantly be at the cutting edge of Deep Learning & Neural Networks – but the libraries within Python or R can do so much heavy lifting “under the hood” to the extent that you can run a basic Machine Learning algorithm with literally just a few lines of code.  The danger there, of course, is that the user should have a good understanding of the mathematical / statistical underpinnings of the algorithm to make sure that the model is valid – so, I do agree that a mathematical / statistical background is necessary… but I’m not sure that it needs to be “extensive”.

I really don’t know your background but it’s not easy to learn to Code or learn statistics on your own. If it was most people would be doing it and developers wouldn’t be paid like they are paid today. Also if by “Data Science ” you mean some basic stat functions in python, then surely you can become a scientist in 6 months. I don’t think however those sets of skills would make you employable. These courses are just an icing ( even less than icing ) on the cake. 

been taking python classes.  these classes just teach you about the tools.  It is your job to use those tools to solve or achieve your goals.  I think it is basically learning how to use an ax, saw, hammer etc.  Now you gotta use combination of these or all of these with other materials to solve your problem.  It is pretty tough and makes one think quite hard. 

This is why coders make such good money.  Not only is knowing how to physically code is important but also to code with grace and efficiency.   you need to find a way to code with clean and efficient lines to problem solve the problem at hand.

Be yourself. The world worships the original.

infinitybenzo wrote:

been taking python classes.  these classes just teach you about the tools.  It is your job to use those tools to solve or achieve your goals.  I think it is basically learning how to use an ax, saw, hammer etc.  Now you gotta use combination of these or all of these with other materials to solve your problem.  It is pretty tough and makes one think quite hard. 

This is why coders make such good money.  Not only is knowing how to physically code is important but also to code with grace and efficiency.   you need to find a way to code with clean and efficient lines to problem solve the problem at hand.

Atleast you have taken the first step? Where are you taking these classes? Any recommendations on where to start.

Thanks

there are many in person classes in NYC.  I take one in the flatiron district arranged by a community college.

coursehorse connects to great classes.  They have all kinds of classes from cooking to programming for all levels, beginners to advanced.

Be yourself. The world worships the original.

infinitybenzo wrote:

been taking python classes.

Make sure that they cover pandas (dataframe manipulation & summarization) and scikit-learn (where the Machine Learning algorithms are mainly found)… otherwise, I think the class would be of dubious value.

yup NumPy, SciPy, pandas, matplotlib, numba.

Be yourself. The world worships the original.

Good stuff.  If you have a chance, check out seaborn for data visualizations (I think it’s way easier to use than matplotlib – and looks better too).

I wanted to get views about this.

Just talked to a friend who took a data science job and the pay was SUBSTANTIALLY lower (70-80k). Plus i felt that the job itself was “glamorised” middle office- cleaning up datasets  in excel file.

Can you guys share your views about this?

infinitybenzo wrote:

yup NumPy, SciPy, pandas, matplotlib, numba.

So basically i am also now taking a Python course (but its not aimed at Data Science). Its more basic programming for practice and basics.

I wanted to ask you if you think I should just pickup some data science course next? or is it better to ‘solidify’ my base because right now i just learnt it in 13 hours of facetime over 2 weeks. (but havent had much practice)

I also took courses and read books and tutorials about Python and R. The problem is once you go past the basic stuff it’s hard to keep practicing and improving if you don’t constantly analyze data as you would in a job with real life problems and data. I always end up forgetting most of it till I take another course but I never reach a good level.

ar169 wrote:

infinitybenzo wrote:

yup NumPy, SciPy, pandas, matplotlib, numba.

So basically i am also now taking a Python course (but its not aimed at Data Science). Its more basic programming for practice and basics.

I wanted to ask you if you think I should just pickup some data science course next? or is it better to ‘solidify’ my base because right now i just learnt it in 13 hours of facetime over 2 weeks. (but havent had much practice)

I took basic class first then data science for finance.  The learning curve was steeper for the second class compared to the first but it was also more interesting.  I try to keep up by going on to Youtube..  I have also been trying but not successfully to somehow use python at work - goal for now is to extract data from either 10k or bbg terminal to my excel spreadsheet.  Then on the fun side, do some statistical analysis on the market.

Be yourself. The world worships the original.

krysix wrote:

I also took courses and read books and tutorials about Python and R. The problem is once you go past the basic stuff it’s hard to keep practicing and improving if you don’t constantly analyze data as you would in a job with real life problems and data. I always end up forgetting most of it till I take another course but I never reach a good level.

make clones of sites/software that you like and use, not for profit, but just for the practice

elcoelhon wrote:

krysix wrote:

I also took courses and read books and tutorials about Python and R. The problem is once you go past the basic stuff it’s hard to keep practicing and improving if you don’t constantly analyze data as you would in a job with real life problems and data. I always end up forgetting most of it till I take another course but I never reach a good level.

make clones of sites/software that you like and use, not for profit, but just for the practice

My problem is with data science. I don’t have ideas to analyze data everyday

ar169 wrote:

I wanted to get views about this.

Just talked to a friend who took a data science job and the pay was SUBSTANTIALLY lower (70-80k).

Depends what you’re comparing it too I guess. 70k-80k is above national average, if you’re “good” i.e can talk through your ass without getting caught you’ll get a 100k easy.

If you’re slightly better than good i.e know what algorithm to use for what purpose and the difference b/w a decision tree and a random forest and how to use things like Spark (more efficient R) you’re looking at 140-170k. If you have enough soft skills to be the bridge as well 200k. Start managing people and 200-400k.

It doesn’t compare to Finance obviously where people are getting 400k before 30 but not many things do.

I also fail to understand why Data Science is going to be such a big disruptor to this sector, it’s basically linear algebra except you don’t even need to know th working behind it because one line of code in Python or R does it for you. Quants have been using these techniques and languages in addition to more advanced techniques like Stochastic Modeling for the better part of a decade.

What is lust without devotion, what are feelings without emotion?

infinitybenzo wrote:

dude i mentioned in another thread that I am taking programming classes.  learn now before the wave of programmers and AI come barreling in because by then it will be too late.

thought you were already rollin in dough by the way you smacked down studdmuffin in the other thread wink

What is lust without devotion, what are feelings without emotion?

Belinda wrote:

infinitybenzo wrote:

dude i mentioned in another thread that I am taking programming classes.  learn now before the wave of programmers and AI come barreling in because by then it will be too late.

thought you were already rollin in dough by the way you smacked down studdmuffin in the other thread wink

I currently roll in “ok” amount of dough because I kept going forward in the past starting from high school, to college, and to the present….Why stop now?

Be yourself. The world worships the original.

Well, people who want  to get CFA designation  should consider study computer science instead.

Just because someone can program, or is a “data scientist” / “quant” doesnt mean that you can beat the market. Investing is a relative game, and almost no-one consistently outperforms all the time. Thats why quant funds arent doing well this year, and they as of yet dont have a clear track record of outperforming discretionary managers either. Sure, passive will be larger. Quant maybe (but currently its more of a marketing pitch that performance chasing institutions will pick up). But there will always be space for long-term concentrated public investors with a PE mindset. In a downside scenario discretionary stock picking will be back down to small niche shops with the $100bn+ managers only being passive/beta. But given the fee structure (% based, automatically get rising profits just by aum / market growth) + operating leverage (fixed costs, no incremental capital, can do everything with 1-2 investment staff) this business can continue to be highly lucrative if you enjoy it. Theres a lot of fat in this industry that is long overdue trimming, kinda ridiculous that sell side analysts are still making multiples of most other jobs including tech. Almost any mega-fund cant justify their fees and the sell side is almost entirely useless and bloated. 

* i would also add that most of the quant strategies suck. there are a handful of oligopolies that can find unique true alpha factors (renaissance, two sigma, aqr) and the rest is just data mining nonsense. mostly just sticking to basic factors with sufficiently large cross sectional datasets work best (quality, momentum, value, carry). you dont need to be a uber quant to implement those. vanguard can do it easily, just need the distribution and technological infrastructure. long story short, you dont get into renaissance by being a cfa and then doing programming/stats because its the next hot thing that is on wsj. you have to actually have enough passion to do a 4y phd in these fields to have a real competitive edge that can allow you to make real money. 

entropy229 wrote:

* i would also add that most of the quant strategies suck. there are a handful of oligopolies that can find unique true alpha factors (renaissance, two sigma, aqr) and the rest is just data mining nonsense. mostly just sticking to basic factors with sufficiently large cross sectional datasets work best (quality, momentum, value, carry). you dont need to be a uber quant to implement those. vanguard can do it easily, just need the distribution and technological infrastructure. long story short, you dont get into renaissance by being a cfa and then doing programming/stats because its the next hot thing that is on wsj. you have to actually have enough passion to do a 4y phd in these fields to have a real competitive edge that can allow you to make real money. 

1. errrr MOST active managers suck.

2. Real competitive edge in quant strats come from how cutting edge technology your firm uses not your educational attainment or your codes.

Be yourself. The world worships the original.