43

I recently joined a technology startup company where a key business value-proposition depends on solving certain various problems that no current business competitor solves. There is one particular sub-problem we're looking to offer a commercial solution to - and it has been already solved by an academic team who published a paper describing an algorithmic solution about 15 years ago. To our knowledge, no commercial operation employs their approach. We would like to capitalize on this.

The team-members are still working at the original university and fully contactable.

The problem is the solution as described in the paper is beyond the current academic-ability of the startup's employees - and given time constraints I'm considering the possibility of contacting the team to ask them to implement it for us - as paid professionals, of course.

Is this appropriate? Is there a good way to frame this? Would paying them for their work introduce any kind of conflict-of-interest or other consequences?

Dai
  • 461
  • 4
  • 9
  • 42
    Your proposal is a bit like saying that you want to hire somebody who designs electric guitars to play in your band. Many academic computer scientists aren't particularly good coders, so hiring one to code for you might not be a good move. I would suggest instead that you get the academics to explain the algorithm to your coders so that they can implement it themselves. – David Richerby Feb 27 '16 at 01:18
  • 25
    +1 for "Many academic computer scientists aren't particularly good coders." My employer has contacts with and sponsors research projects at many universities, and I've never seen a software implementation of something developed by a PhD or Postdoc which we didn't have to recode completely to make it into a useful product - and if the recoded version runs 100 or 1000 times faster than the original, that's about par for the course. In any case, you will end up responsible for the bug-fixes etc in your commercial product, so you need to know how it works! – alephzero Feb 27 '16 at 01:33
  • 2
    @alephzero Yep, I think the paper in question had an implementation written in Matlab or something. I guess what I really want is to ask the researcher "plz email me the codez" - just in a polite and respectful way. It can be argued that any published academic research paper must be easily reproducible, and that includes bundling their source-code in their submission to the journal or conference. – Dai Feb 27 '16 at 03:00
  • If you don't fully understand the solution, how certain are you that it solves your problem? Maybe the proposed solution only works under very special cirucumstances that hardly occur in practice. – Akavall Feb 27 '16 at 03:40
  • 2
    I see no problem with any of the other answers, but also, after thirty years in software engineering, I would think that if an algorithm can be described, it shouldn't be hard to find a programmer who can code it. In many shops, the people writing the code are implementing algorithms described by others. Sometimes these others don't know how to code. Sometimes the coders can't design algorithms. Most can do both. – WGroleau Feb 27 '16 at 06:31
  • 1
    To expand on @Akavall's answer - if you pay the academics to implement it, how are you ever going to maintain it? I suggest getting your developers up to speed math-wise so they can understand it, or hiring someone who understands it, can implement it, and can explain it to your developers. Otherwise you're building your business on quick-sand... – Bob Jarvis - Слава Україні Feb 28 '16 at 02:45
  • To add to what @alephzero has said... We have a couple of researchers who work with us as part of a University partnership. They're smart people but absolutely clueless in a business environment. Things that would take a professional developer a month to implement can take 9-12 times as long. Also, expect to get the "perfect" solution, not an acceptable compromise. None of that is inherently bad, but it can hugely impact product development, especially at a startup. – Basic Feb 28 '16 at 03:26
  • 1
    "Many academic computer scientists aren't particularly good coders". Whether that is true or not, you should judge the researcher on his individual skills - not on some shallow prejudgment. – Zulan Feb 28 '16 at 14:27
  • I'd be curious to hear which algorithm it is. – littleO Dec 21 '18 at 01:02

3 Answers3

46

It is definitely appropriate to ask. University policies on outside work for employees vary. But typically it is allowed. There are at least three possible approaches to this.

  • You hire the academic directly, and it is a side job for them.
  • You provide the university with research funding, which is then used to pay to make the implementation. This might have tax benefits, it will make the academic look good, and will allow the university to take a cut of the money.
  • You partner with the university to obtain funding from a third party (probably a government program). Your company pays for the commercial part of the project and the third party provides pure research funds the academics want. This makes everyone look good but can be slow.

The key thing here is that you need to convince the academics that there is a benefit for them to participating. This is a matter of their personal preferences.

Anonymous Physicist
  • 98,828
  • 24
  • 203
  • 351
  • 8
    Excellent response with good view of the options. Note that, albeit the academic themselves may be reluctant to do this type of job, perhaps they have students completing which may like utilising their expertise. So, that's an option #4. – Captain Emacs Feb 26 '16 at 22:25
  • In the US, that 3rd option is (usually) an STTR and is designed explicitly for grants to transition academic resources to commercial technologies usable by government agencies through small businesses. – tpg2114 Feb 26 '16 at 23:48
  • 1
    It may be necessary to go through the University, depending on the policy concerning intellectual property. Here in Australia, my university has priority on any IP created as part of or incidental to my research position. I'm guaranteed fair compensation for any commercialisation of the IP, but the uni is also guaranteed their cut, so this type of work would definitely have to be negotiated via the uni. – Livius Feb 27 '16 at 00:11
  • We do #3 in the UK. We got smart people doing research and coming up with very accurate solutions. Unfortunately, that's nothing like shipping a product where acceptable compromises are the norm. We're glad we did it, but it turned an estimated 3-month internal implementation into a 2-year project by the academics that had the be recoded from scratch at the end. We got something a lot better than if we hadn't had their help - so we're happy - but we're fortunate that we could sell on other features in the meantime. – Basic Feb 28 '16 at 03:29
12

This is definitely something you can ask the researchers, and something that's quite common in many fields. Paying them for their work can create some minor conflicts - there are often guidelines about how much outside work a faculty member can do, rules about hiring grad students, etc. There's also likely some IP concerns surrounding the work. But most universities have offices to set up these kinds of things, and a simple email to the team is probably a good way to get going. There are a couple ways this can happen (there are more, but these are the ones I'm most familiar with):

  • Consulting. You hire the researcher (or a student, etc.) at a negotiated rate as a consultant. Usually this means they can't access university resources while they do their work, but this is often the fastest and most direct way to get something done. The limits here are that it's hard to make a tenure case (for example) out of consulting work, and if the problem needs specialized equipment, it needs to be your specialized equipment (or you need to buy it for them). It's possible, as @AnonymousPhysicist notes, that equipment from the university may be available for a fee. In my experience however, once you start hiring both researchers and their "stuff", the university starts asking why this is a consulting gig.
  • Some sort of work-for-hire contract with the university. This will likely be slightly more expensive and complex, but may be necessary, and is the best way for the faculty member to get "credit" for their work within the university. This is often distinct from giving the researcher a "grant" - at my institution at least, the university is much pickier about IP, publication rights and overhead rates for grants than they are "Here's a problem, we need you to solve it."
  • As mentioned in a comment, there are both funding mechanisms and often resources within the university to facilitate tech-transfer, licensing and spinoffs. This may help if they already have an implementation, and you just need to license it.

The easiest framing is just "Hey Dr. So-and-so, I work for XYZ Incorporated. We're really interested in your work, and think it might have some commercial applications we'd like to talk to you about..."

Just be direct, open and honest. Many researchers like consulting - it's a good way to supplement their income, work on some small but fun projects, etc. Work-for-hire contracts, similarly, are a decent way to get unallocated funds into a lab, or keep a graduate student or lab tech employed.

Fomite
  • 51,973
  • 5
  • 115
  • 229
  • "if the problem needs specialized equipment, it needs to be your specialized equipment (or you need to buy it for them)." Not true, some university facilities are open to private sector users, for a fee. – Anonymous Physicist Feb 27 '16 at 00:48
  • @AnonymousPhysicist Fair. I'll put that in - with a caveat about my experience going down that route. – Fomite Feb 27 '16 at 00:49
9

Previous answers were about how to structure the formal relationship. I'll focus on how to make it be productive, given the "academics can't code" concern from the comments on the question.

The academics in question clearly implemented enough of their method to produce the results shown in their publications. Hopefully, they even still have that code somewhere. It may not be a complete implementation of the theoretical material their papers describe, though - if the experiments they ran didn't need that, it's quite possibly absent or much more likely to be buggy.

So, here's some things you could consider asking for in this partnership:

  • The original code, appropriately licensed from whoever holds rights to it (maybe the individuals, maybe the university, maybe some other funding source)
  • Test data (ditto on licensing)
  • Their knowledge of omissions in the code or test data
  • Knowledge of limitations of generality, performance, applicability to your problem, etc
  • Ideas on how to make their method more applicable to your problem, and research effort to test those ideas
  • References to related work done by them or others that's applicable to your specific problem, or other problems you may have a business interest in
  • Full participation or technical assistance of various people involved in productizing the necessary code
  • Cooperation in seeking funding for the work from various sources - e.g. SBIR/STTR grants from government agencies

For all you know, some of them may actually be excellent developers, and happy to have funding to put their ideas into practice. Ask a lot of questions, keep an open mind about what each side can get from the relationship with the other.

The biggest take away from my answer is that the code in question may not be the most important thing in achieving your desired results, if it's even important at all.

Phil Miller
  • 846
  • 4
  • 9
  • article quality code is usually not production quality, which is what the comment addressed – Thorbjørn Ravn Andersen Feb 28 '16 at 15:12
  • @ThorbjørnRavnAndersen is spot on. I say that as someone who writes scientist-code. Even something as simple as testing the inputs is often neglected when you're in control of the only source of data. – Chris H Feb 29 '16 at 10:12