Are the Obvious Non-Academic Career Options for Astronomers Not As Attractive as in Other Fields?

by Kelle on December 15, 2010

Volatile and Decentralized: Why I’m leaving Harvard
Tenured Harvard Computer Science professor describes why he chose to leave academia for a gig at Google. (via Evegenya Scholnick)

I get to hack all day, working on problems that are orders of magnitude larger and more interesting than I can work on at any university…At Google, I have a much more direct route from idea to execution to impact. I can just sit down and write the code and deploy the system, on more machines than I will ever have access to at a university. I personally find this far more satisfying than the elaborate academic process.

Does astronomy have an equivalent? What’s our Google? Unfortunately, I don’t think we have one. Maybe this is one of the reasons we struggle so much with identifying non-academic jobs: Comp Sci has software development and the internet, Bio Sciences have Pharma, Experimental Physics has R&D. These folks still get to work on similar problems as if they had stayed in Academia. I guess we have NASA? Skimming the Non-Academic Astronomers Network, I don’t see any job that looks like the Shangri-La this Comp Sci describes. Oh well, no one has yet figured out how to monetize studying the heavens and we need to be more creative and put more effort into preparing ourselves and our students for jobs that are available outside the tower.

{ 15 comments… read them below or add one }

1 Jan Kotek December 15, 2010 at 11:17 am

An option may be to work for free on an open source project. But it does not qualify as ‘carreer’


2 John Gizis December 15, 2010 at 2:45 pm

As far as astronomy goes, this is pretty obvious. Surely everyone has had someone question the decision to study astronomy as not a practical career move?

I do think though that talking about comparing being tenured at Harvard to heading a research group at Google is like arguing about whether you’d rather be the star of the top Broadway play or the top Hollywood film. A few people have those choices and no doubt have interesting insights into why they prefer one or the other, but we have to recognize these are dream jobs outside the normal bounds of experience even for hard-working, talented people.


3 Rob Seaman December 15, 2010 at 11:33 pm

Um…you have it backwards. Astronomy is what people like Wayne Rosing leave Google for.


4 KT December 16, 2010 at 12:53 pm

In response to the points John mentioned, which brought up the following for me:

This issue that Kelle is raising is so important for young astronomers. More and more PhDs are being produced but the Academy is stagnating. Therefore, we NEED other options. Yes, we have considered whether astronomy is a practical career move…but at the time we entered grad school, we thought we could make it. But now that hiring freezes are in place and funding like Obama’s stimulus package are going into lower-tier positions such as postdocs, we are wondering what we will do after our 2nd or 3rd postdoc. Astro2010 leadership failed us, as they did not step up to the plate and make any recommendations about balancing the number of PhDs produced with the number of positions out there. Yet in Astronomy it is SO crucial to balance these numbers, since our non-astronomer career options are SO limited. Computer science can produce as many PhDs as they want, since their skills are in demand in industry. However, if we care about each of our students, we would attempt to move toward a Med-school model where grad admissions are limited by the community, in order to balance the number of positions at the top.

I think going to grad school in Astronomy these days is becoming more and more like aiming to be “the star of the top Broadway play or the top Hollywood film”. Those careers are fraught with risk yet the rewards of success are immense and so that is why people still go for it. However, in Astronomy, so many of us are toiling like would-be-actors waitressing in Hollywood, and the end payoff is becoming less and less possible, statistically-speaking. So if Astronomy wants to continue to be propped up by an army of grads and postdocs, there has to be SOME kind of non-Academic career options that still use those years and years of training… It is just plain cruel, otherwise, to accept hundreds of young students into grad school and then lead them on for years, getting obscure training, ending up with few prospects.


5 DS December 16, 2010 at 4:36 pm

Hey, now, KT, I’m sure there’s huge industry demand for being able to write convoluted, poorly-documented IDL code! Wait, google doesn’t run everything in IDL? Nooooo!

So, um, yeah, good points.

6 KT December 16, 2010 at 5:27 pm

Hey DS, have you been looking at my code library? You describe it very well! 😉


7 EB December 16, 2010 at 7:40 pm

So-called “Data science” could be a good match ( The article calls out physicists particularly, but arguably astronomers may have even better data-handling skills. Machine learning, big datasets, and good statistics are only going to get more important in the years ahead, too.

Encouraging students to use versatile, modern languages like Python for their analysis whenever possible will make them more marketable to such employers.


8 Andrea December 17, 2010 at 11:10 am

EB has it nailed exactly. I’m a former astronomer now working as a software developer, and “big data” is exactly what I sold my skills as, and exactly where I’ve developed in the four years since leaving academia. Big data is an incredibly rapidly growing field, as well as a direction that astronomy is increasingly heading in with the rise of large surveys, so this match should continue to be good in the future.

The problem is that academic astronomy is in denial about the job situation, and continues to view people leaving astronomy as individual failures rather than an indication of a systemic issue. As a result, there’s virtually no mentoring about non-academic career paths, or advice on how to develop skills that are useful both in and out of astronomy (like writing your code in modern languages; this includes using standard tools like Hadoop for handling big data rather than doing one-off solutions.)


9 Rob Seaman December 17, 2010 at 11:57 am

Big data, astroinformatics, and similar terms tend to describe things from the outside-in. How can “proper” computer science (and related fields like statistics) “fix” the naive practices of astronomy? Other phrases like Citizen Science and the Virtual Observatory and the “OIR System” focus more on what astronomers bring to the table. There certainly are any number of niches available for trained astronomers in related projects. (“Niches” don’t necessarily correspond to paying permanent jobs, unfortunately.) This potentially includes “gigs @ Google” and other great organizations like Microsoft Research.

There’s nothing new under the sun and you will wait in vain for Academe – especially in astronomy – to change itself to direct students toward non-academic careers. However, absolutely nothing prevents the motivated student from pursuing broadly applicable training in physics, engineering, or computer science. Discussion forums such as this are very helpful, but the motivation has to come from within. Note that the original posting was about a tenured Harvard professor choosing to pursue a new path. This has little to do with the subsequent discussion. This guy had already succeeded in his academic aspirations – and later parlayed it into an outside job. Presumably that means there was a job opening in his wake 🙂

A word of warning that “standard tools” like Hadoop have a way of aging poorly. Rather, learn the algorithms and system architectures underneath the current buzzwords. Papers like Shannon’s 1948 “Mathematical Theory of Communication” ( or Lamport’s “Buridan’s Principle” ( will be ever young. Read books like Numerical Recipes, not just the latest O’Rielly’s (but read those too). Python is not the answer to every question, just like Pascal wasn’t when I went to school.


10 Andrea December 17, 2010 at 3:31 pm

Rob, of course tools age, and showing that you can keep up-to-date is more useful yet. But being current still beats being ten years behind the curve.


11 Gus December 17, 2010 at 4:36 pm

sorry, rob, but I entirely reject your (apparent) assertions that fundamental CS principles or self motivated training within a excruciatingly slow to adopt environment (isn’t that a self negating argument by the way?) are at all helpful to a young astronomer born into the “fourth paradigm” of data science, which by definition is something new under the sun unless you think Clyde Tombaugh’s plate blinking machine is the right tool for finding LSST transients. Oh, that last dependent clause is a setup by the way. 🙂

So, I second (or third) EB/Andrea’s assertions that data science is already (google:quant) a parallel career track for astronomy, and that ain’t a poor man’s excuse. in this regard both the careers of trainees of astronomy AND astronomy as a science will suffer until the field systematically adopts more formal training in these skills.

and rob, I would hope you would not have protested if Andrea had said “mapreduce” instead of hadoop.


12 John Gizis December 17, 2010 at 10:07 pm

@EB — that’s a really helpful comment.


13 Rob Seaman December 17, 2010 at 10:34 pm

I’ve got nothing against Hadoop – and it wasn’t a protest, rather sage advice from from someone born into the second paradigm. Speaking of which, only a subset of science will ever be data intensive. Physical insight arrives in many paradigms all at the same time. That said, I would be spectacularly unsurprised if IRAF outlived Hadoop as it has so many other buzzwords 😉

I would certainly hope that academic programs learn better how to adapt to (or rather, steer) scientific trends. True pedagogical insights, however, are ageless and Socrates and Descartes still have something to teach today’s students. I reject the premise that grad school is about training and that students are “trainees”. Be careful (and very scared) about demanding “formal training”. Rather, the right verbs are things like “mentoring” and “apprenticeship” and “interning”. “Career” is another dangerous word – think rather “calling”. Grad school is a cybernetic mechanism to transmute a student into a colleague. Colleagues control their own destinies.

The real problem here is that during periods when a discipline is rapidly changing, the mentors may be more in the dark than their padawans (see Wookieepedia). Does the faculty grok Hadoop? How then are they to design a systematic curriculum to teach these skills? Is it a surprise that education is a moving target? In the mean time, get your advisor to send you to Java One or “Telescopes from Afar” or somebody’s Hadoopapalooza.

As far as “fundamental CS principles”, think rather just fundamental principles. Read widely because you want to – not as a means to an end. It is not an insight from the fourth paradigm (more like the zeroth) that pure curiosity is the most successful driver for new discoveries (and the resulting glory and success). Self-actualize, don’t actualize somebody else. If academia is a game, a good strategy might well be to swim against the dominant paradigm, whatever its ordinal. Skills in physics and statistics and engineering (and EPO for that matter) are at least as widely (if not richly) salable as CS. Astronomy? Astronomy per se is spectacularly useless – it’s just that like art and music, astronomy simply serves to justify our existence.

Digital blinking certainly does remain a useful algorithm. All workflows are ultimately human mediated no matter how automated they become. The same eternal truths of empiricism and experimental design will continue to undergird each new paradigm lapping at the academic shore.


14 Ben December 20, 2010 at 5:16 am

The grass always looks greener, etc. That is, it is always tempting to look at Field X and say “Man those people have an easy bailout option! So much easier than ours!” But in actuality, unless you are a very well established person like a Harvard CS professor with contacts at Google, or you work in something like petroleum engineering, the path to bailing out is not trivial for any individual person, even if it looks easy on average and from a distance. I was in physics grad school in the mid-90s (which was a massacre, in certain fields) and people were saying the same things about the supposed difficulty of transitioning out of physics.

You can’t really expect academia to train you on an industry standard tool (whatever the industry standard tool is this year, like Hadoop, which I had never heard of). You can want it to give you experience so that you can learn the latest and greatest tool quickly (_if_ you make the effort to get that experience). Basic research works a lot on prototype, one-off systems. It’s like asking why astronomical instruments are built with so much less process control than mass-produced technology.


15 David Bernat March 17, 2011 at 12:07 pm

I suspect that @Ben is correct: it is not feasible for a university to train its students to use industry tools, just as no industry would allow employees to work on material that is not eventually benefitting the bottom line. But through contact with many graduate students, what I hear often and again for universities to train its students to transition into the industry, by providing resources to a) train themselves in their ‘free time’ (har har), b) begin cultivating an awareness of that vague, time-evolving term ‘industry’, and c) assist in the first contact of students and industry. In my experience, I have heard many students express fear that they have no fallback if they don’t land that tenure-track position or postdoc. Not because they want to leave the field, but simply because they feel squeezed by having put all their chips in so early. I also anticipate that when many undergrads balk at astronomy (if not physics) as a career choice, implicit in that reaction is the question “What the hell would I do with an astronomy Ph.D.? Teach K-12?” Many of us know there are better answers to that question (and if there aren’t, there should be). In that respect, I think websites like the Non-Academic Astronomer Network (which I had not seen before) are a great idea. This information opens up graduate school as a wise professional choice for scientifically-minded students who are focused on professional stability in the long term. I think the field has to ask itself the following question: “Does Astronomy want researchers who give 5-10 years dedicated work who expect themselves to leave the field?”

Also, by industry it is sometimes meant “finance/google/consulting” and the other sectors that draw on analytical modeling of phenomena and events. The other is what has absolutely exploded this decade: data-mining. Grocery chains record purchases to find correlations to motivate sales that can attract customers but keep revenue high; baseball statisticians are literally deriving new measurable variables to drive draft picks and trades. In my impression, these two fields have undergone change in the last decade. which impacts students looking for these careers. Both of these industries followed the lead of the finance industry which hired physicists in droves in the late 90s. Finance has saturated. These newer industries are smaller and less concentrated (i.e., there are no Big 7 to apply to anymore), which makes locating and contacting them a bit harder. At the same time, the traditional “Wall Street friendly physicists’ toolkit” of modeling/programming/datamining has risen in prominance across most physical sciences, and the industry has recognized that physicists need not be their only source for these skills. Both of these mean more legwork for the student. On the plus side, there are many jobs out there and these jobs aren’t going away.

But since the readers here are mostly observational astronomers, astrophysicists, and non-astronomer physicists, I am curious how the readers think the skill sets vary among the three groups. Is one group more suited for specific jobs? Are observational astronomers not?


Leave a Comment

Previous post:

Next post: