Desperate for Data Scientists

Sage advice. There is a significant misconception on what a data scientist does and many roles don’t seem to truly bring out their potential. Most get grouped into doing everything around “data”.

Agreed. From what I can tell the data scientists at my company pretty much are running a few SQL queries with simple stats. They also seem to have a high turnover. It looks like they got hired mainly because the company wanted "data science" without knowing what it means.

That's a very good advice. That expectation and reality mismatch causes a lot of problems and that's also well captured in this article.

https://towardsdatascience.com/why-so-many-data-scientists-a...

>(1) If you're going to hop on the 'hot' AI/ML/DS bandwagon, make sure that you actually have 'work' for the highly coveted person you are trying to hire. 'Everyone else is doing it', 'all our clients have data, so selling them data-sciency projects shoudn't be a problem', 'Hire a first DS person, to have them start a brand new DS division' are things that probably are doomed to fail.

I'd add that if you have all these conditions and still want some hope of getting value from your Data Scientists, give your data scientist(s) a period of time (6 - 12 months) where they can discover how they can provide business value. Accept that during that period they may not provide the value you are hoping for. Support them if they recommend changes (e.g. different ways of collecting or processing data)

Lots of middle/upper managers want to hire a team of Data Scientists and want them them to deliver high business value from day one. They want a magic black box to solve their business problems. This results in exactly what you describe - constant fires, games, gimmicks (that look good on PowerPoint), and good people eventually quitting or getting fired.

On the other hand, if you give them 6-12 months to interact with different teams and identify relevant business problems, they'll find ways to provide value. Or, worst case scenario, they'll tell you (maybe by all quitting) that your problems aren't data science problems and the money you are spending on them is best invested elsewhere.

My company opened up a load of AI/ML roles with absolutely no pending projects for them. CTO literally said once they are hired, they will find opportunities.

That's what I am thinking. Any half decent software engineer could probably produce the results they need from a few SQL queries. But "data science" is hot at the moment.

Concur. On a job hunt this year, my several years of DS drew no bites, while my skills in other niches did fine.

Is a masters enough nowadays for the good data science gigs (autonomy) or is a PhD going to command these gigs?

Not every labor shortage problem is solved by adjusting compensation levels upwards.

The compensation has to match the value created.

>If you fit that bill you can probably get a role paying $95-120k just about anywhere in the US. You'll wrangle data, make pipelines, build dashboards and occasionally deploy a model or two when the planets align.

If you expect to pay that little to anyone who knows the above to any degree you will have a bad time.

A number of projects I have joined absolutely look like someone treated them as a way to learn the technology before moving onto better paying jobs, leaving the business with a terrible mess that's largely unfixable without a rewrite. You're better off hiring one developer that knows what they are doing at above market rates than three who don't for the same price.

> Typical data scientist role I see posted is looking for: - SQL - AWS/Azure - Tableau/Power BI/Other BI tools - Python/R - Increasingly Hadoop/Spark

They are looking for programmers.

Unless...

C programming makes you a hardware engineer,

Fortran programming makes you a physicist,

Ada programming makes you a rocket scientist,

and Lisp programming makes you a god.

It depends on how companies staff their data science teams. I’m comfortable with a lead data scientist that oversees “less math” data analysts who build the models.

If you get the high level concepts of stats and linear algebra, it should be enough that someone else with a PhD can help fill the knowledge gap.

The other way to phrase that is "A data scientist is a worse programmer than a typical programmer and a worse statistician than a typical statistician." If you need a programmer, hire a programmer. If you need a statistician, hire a statistician. I wouldn't ever hire someone who took a bootcamp that is "a better plumber than a typical doctor and a better doctor than a typical plumber."

where do you find php contracting for that which lasts long enough to make 350 a year?

Care to share a link to your licensed ML product, for those of us who're curious? Got any tips for people who are playing with the idea of doing something similar?

> I’d have a hard time finding a job in ML for $300K

If you have 10 years of experience practicing ML, AmaGoogFaceSoft would be falling over themselves to pay you $300k+ in total compensation.

Whether you'd want to move to SF/NY/Seattle and work for them, though, is another question!

IMO, as companies' data science capabilities mature, people with actual skills are needed.

5 years ago DS was hot and new. As a middle manager you would call up a consulting meat farm and ask for data science consultants: presto, you have a data science team. These people would, likely, be random BSc's with a SAS base certificate.

In 2018 data science, and data driven decision making, is still hot, but hiring managers know they need people with a formal education. So, education reqs go up and scarcity increases.

In my experience, data science is turning massive amounts of potentially unstructured data into some sort of insights for business. I guess depending on the volume and structure of the data, and the level of insight required, you may need to be more or less qualified to do the job. A bachelors degree would be fine for most of the roles I’ve seen in places I’ve worked, but I’ve never worked anywhere that had google level demands.

I think getting insights out of massive amounts of unstructured data is a relatively new problem for most organisations, so that might explain the skill gap. Personally I also think the ubiquity of ORMs might also play a role, as they remove a lot of the day-to-day data handling that many engineers would otherwise have to do. But I might be completely wrong about that.

I think we are still in the wild west of the Data Science field, at least in the Midwestern U.S. I have seen job postings for Data scientists that are simply rebranded data analysts to postings looking for advanced math and programming ability.

I am suspicious that bootcamps and online courses will set individuals on to long term data science careers. I think there will be a large demand for individuals who understand basic data science principles such as the difference between supervised and unsupervised learning problems. I think these people will be given subscriptions to services that automate a lot of the actual data science work flow. They will use programs like Data Robot. However, I also think these individuals will be paid closer to what a data analyst is than what a software engineer gets.

I think that certain companies will always be looking for people who can actually do traditional data science work from scratch because the automated tools will not apply to their specific use cases. I think these individuals will probably generally have a master's or PhD.

I think right now the field is too new for managers to be able to tell the difference between what kind of Data Scientists they need. My experience has been a lot of companies need the former, but they think they need (or they want) the latter. I suspect in the coming decade these two roles will find different titles.

There is no gap, actually. It is plain, old supply vs demand. Second and third class firms wanting the few Einsteins out there for peanuts, being flooded with CVs from kaggle noobs and mooc kids. Point is the latter are good enough for the jobs on offer but the former feel somewhat entitled to be choosy & stingy, even if they have no real business case to pursue.

There is some value in the additional rigorous data processing skill set.

The intent is to get away from just merely slicing and dicing data to come up with some insight that explains causality.

Whether we are afforded the time to do a proper and thorough analysis is a different matter.

I'm on a hiring committee for a job with the title of "Machine Learning Engineer". The most qualified candidates we have come through the process are physics PhDs who did a significant amount of coding during their PhD and have been able to segue into machine learning because they were interested.

Our other tier of hireable candidates have been individuals with 4+ years of industry experience, usually with CS PhDs with a machine learning specialty.

The rest of our team came from internal transfers, people who were less qualifed but proved that they would have a positive impact on the output of the team.

As the hiring manager for that kind of profile for more than ten years, the problem is rather simple: they want data scientist, but they don’t have good data. The position are still open but no one well-informed would apply.

Managers want to understand what’s happening in their company; they’ve been told data scientist can make that happen, or better, fix the problems automatically. They expect the solution to come without significant investment in data collection. One, even over-paid, data scientist is a far smaller budget than having your dozens of teams make an effort in fixing data collection, and setting up a reporting team. That kind of problem is easy to detect during the interview, if you’ve been there, and any data scientist has. What’s left are fresh graduates without diploma, who were taught statistics, but not how to get a project running -- so they don’t know what to say during an interview.

In more details:

Too much focus on stats. The actual skillset required should not be that demanding on statistical side: interviews generally focus on testing that exclusively. It takes the form of Whomever has any statistical background will grill on what they know best, their own MSc thesis. The thing is: no one can know the minutia of every model, certainly not a student’s interpretation of them, so those end up being a lot of “I would have to go back to the documentation.” Unless your interviewer is mature enough to either know that’s reasonable or stop doing that, it’s a lost cause. At best, you end up hiring a pile of people with the same focus, and every problem will look like a nail to a group of people carrying hammers. One alternative is to ask about something they did recently, but that’s rarely very instructive, either because the interviewer has no idea if that makes sense, or because the implementation has a lot of properties that make the whole process tedious. The focus should be the ability to learn new techniques at speed. Because the vast majority of the time, the right model is:

a. a basic implementation of a better one, just to get things started (fix the data collection, the engineering of the input to the serving model, think about scaling the serving model, improve the actions taken from the decision, improve the decision format and the actual model objective)

b. a model that you haven’t heard about, and you need to learn about it within a calendar that is conditioned by a. not being sufficient.

Be more project-focused Description like “codes and does statistics better than either” fail to mention that the bulk of the time, a data scientist is actually more of a project manager, trying to identify blockers, get the integration of their tool prioritised, compare the impact and risk of different approaches; there’s also a big chunk (if not all of it) of being an analyst, generally to clarify what problem is being asked, or checking that the data is consistent. That’s if all is reasonably good: far too often, the time is spent trying to get management to grasp that Machine learning can only predict something, and that what to do with that, what to expose to the users, how, when, is more the remit of the product team i.e. you don’t have a real direction. Many large companies actively select against people willing to say that the kind is naked.

Some company test for “business sensitivity” but that too often means checking that aspiring data scientist can walk through a P&L. Reality is closer to: Can you listen to a project manager, frustratingly oblivious to his own contradiction and general ass-holery, rant for 45 minutes without strangling him? Because you need the support of his team to get through.

Fight for good data I have yet to see a company that had data you could use for data science on day 1. The vast majority of the time, it was inconsistent. Not a little bit, or hard to find: big, glaring, obvious contradictions like 20% of the revenue unaccounted for. If there are analysts, they generally have found a way around it, rather than address the issue. In more mature companies, there were some problem that would make basic model stumble, and more nuanced one go to very stupid conclusions. The classic case of a trip-up on good, but not well-structured: What drives retention? Completely failing to delivery your service: people will re-order within the hour, a larger ratio than any other case. Executives pay for the servers so they think that data quality if measured in peta-bytes. They get fed a continuous streamed of over-analysed data, filtered by hand from any inconsistency. They have no idea how inconsistent it really is. Any improvement is seen as a cost centre, rather than the ability for the company to be better informed.

Don’t get me started on technical debt, short-term focus, insensitivity.

What data scientists need, before they even join: clear objectives, clear breakdowns to understand what effects are correlated, clear plans to address some of the concern, pilot tests to make sure that those solutions would work, and collect some data, audited data pipeline with clear documentation to avoid misinterpretation, dedicated engineering to help with data logging, hosting computation, serving model. I am willing to bet that hardly any of those dozen of thousands of jobs looking for candidates don’t have most of that.

There is a small piece of good news: Agrawal & al. summarised their theory on how ML will make predictions cheaper in a all-access book, Prediction Machines. The predictable success of that book will help. Basically, it tells about what ML does, without ever doing any ML at all: it just looks at how it’s a lynchpin technology, like search engine, digital photo, etc.

Most data scientists I know don't have CS degrees but instead a background in physics, biology, math, economics or statistics.

For a data science role focused around exploratory analysis, model building and idea generation (as opposed to a more engineering-focused data scientist) I'd say that a basic understanding of programming concepts and familiarity with the most essential tools (version control, Linux, Python/R) is probably enough to be accepted into a data science role. Data science degree programs are still quite novel so most companies don't expect to recruit people with such a degree, hence there are good opportunities for people with a STEM education I'd say.

I have a degree in mechanical engineering and 2 years later am a software engineer pivoting into a datascience/dataengineering role after building an nlp product for our business.

I only have my experience and don't know how repeatable it is, but I just got here by building stuff that I thought was interesting.

Right at the end of college I tried to build a connected hardware/data harvesting startup, and it failed because I was a naive idiot who didn't understand just how hard the financials of building a hardware company were (all of the money up front for manufacturing runs) and how much time I would have to spend on things that I wasn't good at or really interested in, like building out supply chain and design for manufacturing or dealing with radio emission problems.

But over that time I collected a couple of relevant awards (decently sized pitch contests, won at a mlh hackathon I did for fun etc.), built interesting prototypes of the hardware that worked and a web platform that supported it that people thought was interesting and that all worked as a hedge to show that I can at least build interesting things. So when I looked a recruiter came to me with a software engineering job at a startup that I thought was novel and interesting and I ended up taking it.

Once I was there I saw a problem with a lot of labor going into dealing with relations between a bunch of types of texts, so I built a preliminary thing to automate as much of that as possible using relatively basic clustering and putting a person in the loop to collect labels so that I can come back and try something supervised later. The CEO brought on a chief data scientist to help me fix everything that I messed up and build a more scalable deployment pipeline and architecture for datascience microservices and now I work with him.

But yeah, my experience is just that I built things and tried to make sure they worked and were sane, and then just showed people those things to convince them to let me do that for other things I wanted to do. I have a couple friends who did relatively similar things with pretty similar results.

I’d be more inclined to think that at a startup, because of the lack of infrastructure, a data scientist will do a lot more than research.

>Companies doing ML/AI will usually have a small team of Research Scientists who mostly hold PhDs

How many companies are doing genuine AI research, as opposed to applying the research and tooling to their unique business problems?

The UG online masters of computer science is a good option. Many people in this site speak highly of the curriculum[1]. Last time I checked the total for the entire curricula if you pass every course on the first try is around $7000. As a bachelor holder myself looking to break into some of these higher-salary and in-depth roles I’m certainly considering it. It’s even better if you have a company that will pay for it, and because it’s so cheap even the most meager offerings from companies will cover a good portion of it.

1: https://news.ycombinator.com/item?id=15018002

If your target is a data scientist role at Google, you're probably going to want more schooling.

But if you've been producing value in a DS/ML role for years, you have experience, which is even more rare than some of the qualifications people are listing here.

If you can say "I created an anomaly detection system using isolation forests that 5,000 clients relied on for detecting market changes", there will always be places that want your skillset: it is kind of a new field, after all.

So the credential bloat at entry-level shouldn't really be an issue for you.

Maybe a bit (a lot?) out there: why not go for a business degree instead, e.g. EMBA?

You mention “foreign country” so I’m guessing you can have access to good curriculums without paying the US premium on education (or just go for an online course).

Pros: the time and money is not “wasted” as you actually pick up new skills; the MBA card should be enough to trump any education requirement; and you become that most desirable of hybrids: the tech/data guy who can talk business (or vice versa).

Cons: significant time (and money) investment; doesn’t help you get expert DS jobs (you’d be aiming for team/program manager, consultant, etc)

My own experience: completed an EMBA in 2017. Ranked in FT’s top 10, the program cost was around 50 k€ (it’s increased a bit since) and I was able to get 25 k€ of outside funding. The program I followed lasts 2.5 years, meaning I was able to do it while keeping my job (and having a kid) without losing my sanity or my wife. Landed my dream job just before completing the curriculum for a nice 40% pay increase (not saying the EMBA alone had that effect, far from it —but it definitely helped).

>As someone who has been producing value in a data science/machine learning role for multiple years, it's disheartening to see comments that I may be blacklisted from positions due to "only" having a bachelor's degree.

Don't worry.

The big salaries will go to people who create value and solve problems. You can do that without a PhD. In fact, if most Data Science communities are representative, PhDs feel they're above 90% of the work required to put data to work to solve problems. You know, the ones who walk into a job and say, "Oh, I don't get to apply the latest algorithm onto a perfectly cleaned toy data set? I'm leaving!". They're going to have their lunch eaten.

>Any advice on how I can effectively jump off the black-lists?

Find a decent hiring manager who has actually done some hands on Data Science/Analytics work and knows what skills/thinking are actually required. Lots of Data Science hiring managers have no or limited practical experience with doing actual Analytic work so they get overly focused on paper qualifications and buzzwords. This is reinforced by HR people who love buzzword bingo.

I created a vehicle plate recognition system before ML got cool, but I can't get any ML job with my less than bachelor degree (associate? dunno how to translate) here in Brazil. I think there are only 2 positions open for machine learning less than 100km from where I live.

I feel left out.

Errm - I am pretty sure that we didn't have MCMC or causality analysis in the 90's.

Or R-studio.

The problem with stats degrees is twofold:

1) They often lack experience with messy (i.e. real) data 2) They sometimes lack experience with programming.

Data science, as currently practiced requires some CS, some stats (with a focus on experimental design/observational analysis) and communication skills.

These skills are rarely taught together in undergrad, but lots of PhD's acquire them by accident, hence the propensity of PhD's to end up in these roles.

It's highly unlikely. There is a lot of competition for entry-level jobs and not many remote positions available. The barriers to entry are exactly academic qualifications over certifications and portfolios.

Remote job? Probably not without experience and/or connections.

Portfolio is going to be the most helpful, followed by certifications and nano degrees.

That’s why we usually have a team that complements each others’ skill set. The data scientist while great to have everything does not necessarily need to have everything. I tend to focus on their core strengths and depending on what their growth/career aspirations are, help to develop the skills. If it’s soft skills, happy to help them develop that edge.

Data scientist buzzword compliance:

Spark, Hadoop, Hive, Pig, SQL, Neo4J, MySQL, Python, R, Scala, Tensorflow, A/B Testing, NLP, anything machine learning

Masters and PhD are required for particular employers. Others not so much but there are caveats. High probability of being the “data” guy who is then made to be the database admin.

Better summary at: https://blog.insightdatascience.com/from-phd-to-data-scienti...

My experience is that most valuable data scientists are those who are good engineers.

Yeah.

I find "data science" a cringey expression and I'm not looking for a job currently, so my LinkedIn emphasizes my strengths according to my ego/self-appraisal, not a "desperate" reaching out for fashionable job titles.