Friday, August 23, 2013

Ask CJ: Computational chemistry jobs?

A reader writes in with a question about computational/theoretical chemistry jobs:
I was curious as to what the CJ readership has to say about computational chemistry, it's prospects in industry and academia, and any general thoughts on computational/theoretical chemistry.  
Personally, I went into this field because I fell in love with quantum chemistry in my undergrad. The complexities and vastness of quantum essentially boiled down to our desire to learn how things work. I mean really work. Computational chemistry was just the applied application of that. Recently I've found that computational chemistry is being used more and more to compliment all types of chemistry and with the increase of computing power it has really boomed in popularity in recent years. Not only is it more user friendly, but its much more accessible now as well. 
So here's what I've said about computational chemistry in industry, here, here and here. Wavefunction also wrote a very worthwhile post about the topic, which he knows a lot about. I concur with one of Wavefunction's original comments: "Comp chem has always been a niche slot, with low demand and low supply."

But here's something I will say that's positive about the field -- if you decide to leave chemistry for another field, you could probably do worse than to have a lot of computing/coding experience.

Readers, you probably know a lot more than I do about this -- what do you think? 


  1. A bit off-topic: I am in 1/3 a synthetic organic chemist, 1/3 physical chemist and 1/3 computational chemist. This combination of skills helped me a lot to get me my current job (r&d in a large corporation), as well as the previous one (postdoc in a middle-sized uni in western europe). But I believe that 'pure' computational chemistry is a niche thing...

  2. The coding skills that a typical computational chemist learns do not make them software engineers, so these skills are not as transferable as they might seem. Imagine the poor computational chemist at an interview for a serious coding job when the interviewers ask what languages the candidate is familiar with and he responds "FORTRAN!". After a few minutes of the interviewers laughing hysterically, the regain their composure and say "No, seriously, what languages do you use?"

    Revision control, error handling, algorithm design. These are the things that computational chemists typically don't concern themselves with. In a world where transferable skills were in demand, this wouldn't be a barrier, but we live in a time when employers are looking for purple squirrels.

    The prospects for computational chemists working in the traditional bastion of pharma get bleaker every year. My advice for the computational chemist looking for work in industry is to familiarize yourself with statistical tools such as SAS and R and a more general scripting language like python. Then they can work at a bank or marketing agency. Analyzing large data sets is a marketable skill that computational chemists typically do have. Time to cash in on that "Big Data" hype.

    If academia is your thing, enjoy postdoc limbo, because there is a glut of qualified computational chemists, especially from top programs that are really just PhD factories.

  3. It's hard to contribute meaningfully to this topic without writing something ridiculously long (which I don't want to do, and few will want to read). Moreover, every story is an anecdote unless you're a labor economist. With those caveats, though, I'll try to offer my possibly informative anecdotes.

    Over 21 years, I've put an average of about 2.5 grad students or postdocs (modelers/theoreticians all) onto the job market per year. They all found gainful employment (either an immediate offer of some sort, or an immediate postdoc and subsequent employment). A scan of the list includes, on the industry side, staff scientists at Kodak, Dow, Proctor & Gamble, Johnson & Johnson, Merck, Pfizer, Samsung, Agouron, Cray, Accelrys, and Schroedinger. Several former co-workers started to and or continue to work in government research organizations (NIH, Singapore National, NIST) -- note: NRC postdoctoral fellowships are FABulous opportunities in that regard -- I'm always shocked at how few people seem to know they exist. The remainder are in academia, ranging from one at a Quaker high school to the vast majority at PUIs, to two or three at Masters/Ph.D.-granting institutions.

    Are there secrets? Nothing people haven't heard, I suppose. Good networking started early. For those interested in academia, Minnesota has a strong Preparing Future Faculty Program that my co-workers have almost always taken advantage of and recruiting schools almost always have been pleased to see. The NRC postdoc option I've already mentioned as a great way to get a foot into national laboratories.

    For industry, a point well emphasized in @CuriousWavefunction's highlighted post above is that you simply MUST go to an interview prepared to speak the language of your experimental colleagues first, and your (possible) modeler colleagues second. At the end of the day, almost all companies MAKE things, and theory plays a service role -- maybe a really CRITICAL one (Merck modelers went to the White House when that company's AIDS drug won a Presidential innovation award...), but it's still service to ultimate production.

    Now, if you ask me about the future, I'll spare you the apocryphal Yogi Berra quote and say that MY opinion is that the next high demand area will be for theoreticians well trained in the solid state -- materials, semi-conductors, batteries, fuel cells, zeolites, MOFs, heterogeneous catalysis -- that's what looks hot to me just at the moment. As noted above, big pharma is at best pretty steady-state right now for modelers and I don't see that changing much anytime soon. "Big Data" jocks (or jockettes) will certainly also find their skills in demand, but that starts to shade outside of what I'd call Chemistry and more into a generalized skill crossing many disciplines.

    I will note that most federal agencies these days like to fund "team"-type grants more than they do single-PI grants, and most teams WILL be told that they need a theoretician/modeler if they don't have one. So, if you LAND a job in academia, you won't lack for opportunities to collaborate. But, that's a post-job observation, not a job-hunt help (unless the help be to note that while interviewing you should emphasize your interest in a judicious mixture of collaborative and individual projects).

  4. > with the increase of computing power it has really boomed in popularity in recent years.

    And they can come up with the wrong answer SO much faster! gr

    I really think a lot of scientists would do well to take classes in Numerical Analysis and Computer Organization (primarily so that they understand computer arithmetic).

    My first real job was systems engineering (Military space - the market for chemists was really bad at the time - sound familiar?) The fact that I did really well in fixing Other Peoples Mistakes got my career started (and I haven't been back in Chemistry since, but I've tried to keep up a little on the subject).

  5. Computational chemist checking in. Got my PhD with a recognized name in the field and well-regarded fellowship to take me through my postdoc, and then... nothing. I relocated to one of the biotech hubs and looked for industry positions for a couple years with no results. I worked some temp and temp-to-hire positions that had nothing at all to do with computational chemistry. I obviously could have timed my exit from academia a little better, but from my perspective there is very little to no demand for "fresh" computational chemists in industry. Tried to transition to silicon valley tech, but those companies seemed even less interested in a 30+ year old with chemistry degrees, however talented I was with a computer.

    I am now working in government in an area well outside what I trained to do, and I love it. I look forward to work everyday. I probably couldn't have landed the job without my degrees, so grad school served its purpose. I wouldn't advise anyone else to follow in my footsteps though. The feds aren't exactly on a hiring spree.

  6. I concur with Chris Cramer.

    My students and postdocs have done well because theoretical and computational chemistry is a twofer. On the one hand, they know physical chemistry very well, can solve many useful problems on the "back of an envelope," and they know how to frame complex chemistry programs in mathematical frameworks that don't always admit to analytic solutions. On the other hand, they can develop algorithms, write codes, and use black-box codes to solve real-world problems. As Chris also remarked, the ability to communicate with experimentalists and technologists is critical to their success. With such training, there is little wonder that they have gotten jobs in academia or industry.

    Two of the quotes in the original post deserve some additional clarification:

    (1) "I went into this field because I fell in love with quantum chemistry in my undergrad. [...]"
    This suggests that computational chemistry is the same as quantum chemistry. In turn, many equate quantum chemistry to the determination of electronic structure. Both of these are in error. In chemical problems, quantum mechanics enters in describing electronic structure, but it also enters in describing the nuclear motion in a number of ways. Meanwhile, dynamics and statistical mechanics are important branches of computational chemistry. Indeed, many applications in computational chemistry—e.g., the structure and motion of polymeric systems, self-assembly of (nano)materials, and protein folding and binding—rely primarily on nonequilibrium molecular dynamics methods. While electronic structure is also an important branch of computational chemistry, it's important to recognize that the latter has much greater scope and thereby impacts a broad swath of chemistry.

    (2) "Comp chem has always been a niche slot, with low demand and low supply."
    While this quote is funny and likely written intentionally to be provocative, it is far from truthful. In chemistry departments, roughly 10% of the faculty are card-carrying theoretical chemists, and an even larger percentage use computational chemistry tools routinely to advance their research programs. At Georgia Tech, our theoretical and computation chemistry groups include five professors, and well over 50 students, postdocs and research scientists. That's just the ones in chemistry. The list doesn't include the many students involved n the theoretical and computational chemistry efforts in the schools of Chemical and Biochemical Engineering, Physics, Biology, Materials Science and Engineering, and a few others. All of these students will get jobs. Thus computational chemistry is far from a niche field!

  7. The view form the top sounds pretty good. Not surprisingly, both tenured faculty who chimed in have a rosy picture of the payoff for a computational chemistry degree. I suspect lottery winners are bullish on lottery tickets as well. I'll submit another observation from the trenches, and it's not so pleasant down here at the bottom. PIs are usually quick to claim any former student who isn't starving in the streets as a success, even if their degree had nothing to do with their ultimate career (see the anonymous computational chemist above). Multiple postdocs after graduating and eventually losing touch = successful mentoring outcome listed in the CV. I also seem to know an inordinate number of world-class computational chemists teaching high school chemistry. Sorry, that's like going through Navy Seal training and getting a job as a mall cop. Unfortunately for them, they can't take the path of Walter White and leave high school to become a drug kingpin, because the meth superlabs aren't interested in QSARs.

    Secondly, I submit Rig's pedantic nit picks above as evidence of just how out of touch academics are with the real world. Using Wavefunction's statement "I went into this field because I fell in love with quantum chemistry in my undergrad" as a excuse to get on his hobbyhorse should tell aspiring grad students everything they need to know. And as for his prediction that "All of these students will get jobs." - High school chemistry, here I come!

    Finally, these outcomes are only for those students lucky enough to finish their PhD (or desparate enough to stick it out. I always think of the movie Officer and a Gentleman when the drill instructor pushes private Mayo to D.O.R. and he says "I got nowhere else to go!"). Think about that prospective grad students. There is a decent chance you will be fired or you will quit. Those people outnumbered those of us who graduated with a PhD in my group.

    To sum up: computational chemistry faculty say the world is your oyster and things are looking bright, and recent PhD asks if you have any spare change.

    1. The person who loved quantum chemistry was the person who wrote in, not Wavefunction.

  8. Pedantic nit pics they may be. However, it's useful to understand that computational chemistry is a broader field than electronic structure. It's true that the latter is sufficient to solve many problems in chemistry, and it's not surprising that the anonymous writer fell in love with quantum chemistry. It's equally important to recognize that the broader field of computational chemistry is available to solve other problems too.

    As for the definition of success... It is absolutely true that very few chemistry Ph.D.'s, regardless of discipline, become faculty members at research I institutions. If this were the only (or even the primary) reason for doctoral programs, then the system would be broken from the beginning. I aim to train my students to be successful in the broad set of positions that are available to them in the marketplace.

    The market for jobs is tough now (and has been for some time) regardless of discipline. While fluctuations exist (sadly some in the negative direction), on the whole, theoretical and computational chemistry degrees provide individuals with a large number of good options as they continue their career. I base this on an aggregated cohort of the graduates from my program. That's still a small sample. It would be useful if a larger study were available on this question, but I don't think that would change the conclusion. Namely, that theoretical and computational chemistry degrees are comparable (if not a little above average) to other doctoral chemistry degrees in subsequent career advancement and satisfaction. These are averaged outcomes so the claim can't include a guarantee of mad money or prevent joblessness for particular individuals.

  9. What's the long term prospects of theoretical/computational chemistry in the U.S./Western Europe/Japan versus the BRIC nations? It seems like the outsourcing that software/IT has experienced could easily be extended to this particular sub-field.

  10. Late to the party (blame a broken foot) but Chris sums up my views better than I ever could so there's not much to add. It's clear that Chris has done a really terrific job of training his students to appreciate the experimental and physical aspects of chemical systems; that's indeed one of the most important skills a comp chemist brings to the table. Although my job description requires me to be a modeler, I also put on the hat of crystallographer, biochemist and physical organic chemist (that last one quite often), and I am appreciated all the more for this broad understanding. While getting a feel of chemical systems - things like sizes of atoms, hydrogen bonding propensities, conformation and sterics - is important for chemists in general, it's particularly so for modelers. In any case, the take home message is that if you are a comp chemist, you need to bring much more than working familiarity with all kinds of software and programming skills to the table.

    I would also like to address anon 10:15's question. In theory you can imagine modeling being outsourced if all that it required was the ability to write scripts, push buttons to dock molecules and run MD simulations. But as my description above indicates, it's more than that. IMO you can easily find someone who can follow computational protocols, but it's much harder to find someone who has an intuitive sense of the system, can judiciously pick approaches based on their strengths and limitations and get a physical feel for what the real results mean within a margin of error and based on experience and familiarity with other similar systems. I am not saying such people don't exist elsewhere, just that it's harder to outsource that kind of understanding and qualitative analysis.

  11. As computing power drops and powerful open source software is available for computational chem the subject is no longer being regarded as niche. PhD chemists are beginning to turn up at interviews with this software on their computers and some ability to use it; as this happens more and more it is becoming expected.

  12. Another computational theory chemist here.
    1) generally a lot of the code used in the fields I worked in involved fortran and some C. While this has made it easy for me to pick up python, I still did not have the speed or programming skills that a computer science engineer or a commercial software developer would have. My programming skills were secondary to learning the physics and chemistry that I needed to know. And in the job market, plenty of engineers and physicists also have a smattering of coding experience. So "transferable skills" to a coding job is not that likely. You would need to be very very lucky.
    2) big companies can afford to buy black box software for their experimentalists to use, they don't necessarily want or need to hire a computational scientists.
    3) theory and computation are cheaper for an academic setting than buying zillion dollar equipment for things like material science, laser science etc. So universities in countries like India (large population=larger talent pool) that might not have the funds some areas of research still often have very strong computational groups. Combine this with the fact that computer results are easy to share long distance, and you see many multinational companies hiring lots of computational researchers... in India.
    4) many of the industry jobs that you could do or would like to apply for might actually ask for an engineering or physics degree. Be prepared to have automatic application screening software immediately eliminate you if your degree is in chemistry.