In this episode, I interview Dr. Nathan Fulton, a philosopher-turned-semantic architect currently working at Indeed.com.
Books mentioned in the podcast:
You can email the podcast here: firstname.lastname@example.org
Also on Twitter: @OffCampusPhDs
Music credit: Robert John - Changes
00:51 Grad School
08:47 Transitioning into Alt-Ac
20:02 Ontologist at Cycorp
29:21 Semantic Architect at Indeed
39:10 Learning More About This Path
46:43 General Advice on Alt-Ac
Kino: You are listening to Off Campus, a podcast about humanity scholars in the alt tech world. I'm your host, Dr. Kino Zhao. This is episode two, where I interviewed Dr. Nathan Fulton, a philosopher who currently works as a semantic ontologist. In this episode, I ask Dr. Fulton what exactly a semantic ontologist is and why philosophical training seems to be especially relevant for this career path.
Kino: Dr. Nathan Fulton received his PhD in philosophy from the University of California, Irvine, specializing in game and decision theory and political philosophy. He then joined Cycorp as an ontologist and programmer. Nathan currently works at Indeed.com as a semantic architect.
Kino: For those who don't know. Nathan and I received our PhDs from the same university, albeit from different departments. So, Nathan, can you tell me a little bit about why you decided to do a philosophy PhD in the first place?
Dr. Nathan Fulton: Well, when you have a severe case of the philosophy disease and it's a chronic case, then you you sort of have two options. You can try to just manage the symptoms and see if you can reduce them and live a normal life. And sometimes that works. And what I mean by if you have the philosophy disease, I tend to mean if you're attracted to and tend to want on your own to talk about, but also sort of seek out conversations with other people about pretty fundamental questions about things like what can we know and how can we know it? What is the what sorts of things are there in the world? What's the best way for a human to live? Stuff like that. The really core questions of philosophy. If you find that you just can't leave those questions alone, you're often better off finding a way to build a significant part of your life around that so that instead of having it get in the way, it's helping you find meaningful work to do, or maybe even just be a benefit to other people who are attracted to those sorts of questions. So I in general was pretty aware pretty much from the time I switched to a philosophy major that I was going to go and want to pursue graduate level education in it. And I originally wanted to go to UC Irvine in particular in order to work with Aaron James, who who I was fortunate to have as my doctoral advisor, because I'd seen him speak before. I'd read some of his work, and I just thought he was asking some questions that were important to me as well. But then additionally, having the Department of Philosophy, as you sort of alluded to, we have our sister department, the Logic and Philosophy of Science Department, which is just an incredible resource with just really some of the finest minds in the world working on some key topics. So I was just very lucky to have access to that as well and to be able to do work connected to both departments and and bring that together.
Kino: So was your philosophical itch within political philosophy? Can you tell me a little bit more about your the kind of research you did during graduate school?
Dr. Nathan Fulton: For me, a lot of the most interesting, and not just interesting, but also tractable questions that I could actually see a way to get at them. A lot of the most interesting, intractable questions were economic ones, and in a way, a lot of what I was doing was sort of what used to be called political economy that goes along with political philosophy, but canonically. You might think about political philosophy as asking questions like What are our fundamental rights? Or Where does the source of political authority come from? That sort of thing. And that wasn't quite the approach that I was interested in. I was interested in asking a question that was something more like, what can we do together? What can we do collectively in a group of a group of people who are sharing resources and who are agreeing to together abide to a certain system of rules and that sort of thing. Even in cases where we're not family, we don't necessarily all even know each other. So I approached those questions through the lens of economic game theory. But of course, economic game theory and evolutionary game theory and other kind of areas of inquiry and methodologies within that for modeling these things, they're all very intertwined. So those were the sorts of things that I was interested in, questions about things like if as a society we have to make policy decisions about managing risk. What can game theory show us about that? And then for my dissertation, I was concerned with taking a view that's actually now a very traditional view in political philosophy, the libertarian or more specifically, the monarchist view that the view that what we want to do is try and figure out what's the smallest possible.
Dr. Nathan Fulton: A whole state entity that functions well politically. I wanted to look at that view because it has clear, stripped down elements. Not only is it a view about state that has been stripped down only what it requires, the theory itself tends to be stripped down to some really basic requirements like that. We have a right to freedom of association, to cooperate with other people and exchange or share resources and set up rules and do that sort of thing. We have the right to engage in collective enterprises like that of whatever kind we want, and also to withdraw from enterprises like that provided that we don't have any outstanding debts or obligations. So that's one part of it. And then a particular view of property that's really expansive. You can kind of just say as long as they have those characteristics, which to some extent most contemporary states that sort of have the heritage of the Enlightenment kind of have those characteristics. They have at least a little bit of a libertarian strain to them. Then they kind of fell under the kind of thing that I was modeling. So my dissertation examines what are the strengths and weaknesses of that way of approaching things. If we think about it from the point of view of will, if we look at it, we model it in using the tools of game theory. Is such a state going to be stable over time? Can those is there are there tensions between those two kinds of rights, the property rights and the freedom of association rights and that sort of thing?
Kino: Other than the specific research topic that you pursued, what was your graduate school experience like overall? Did you enjoy teaching?
Dr. Nathan Fulton: The teaching quite a bit. I mean, I love graduate seminars, so still certainly kind of miss being in graduate seminars. Those two things, actually. Participating in these kind of these discussions that are an academic quarter long sort of ongoing discussions among a tight group of people who are really interested in topic and have a lot of a lot of background in it are really diving into the fine points. I missed that a great deal. And I also miss teaching. I think having to occasionally reinvent but also repeatedly refine the way that you explain some of these key foundational concepts really helps you understand them, and it gives you sort of grounds to connect to people that you wouldn't really have reason to have in-depth conversations with because undergraduates, some of them are going to be majors. Sure. And those and a lot of the time, hopefully, they're they're going to have a particularly strong level of interest, but also the the non majors who are sort of passing through your department for an elective or maybe to fill a requirement without having to take a math class if they're if they're taking symbolic logic from you. Those conversations are really interesting, too. They just really help you stay grounded in how people express their values. Since the majority of your students are usually pretty young, it helps you retain a sense of how our values are changing. And I've always found that really interesting, really valuable. I was fortunate enough that I got to be the instructor of record towards the end of my time as a graduate student and design some courses and shape them a little differently. And that was also just a great experience.
Transitioning into Alt-Ac
Kino: When did you first think about leaving academia?
Dr. Nathan Fulton: I would say around the time I completed my prospectus and was ABD [All-But-Dissertation] I was sort of thinking about having a plan B and a plan C. Plan A was still for me to have an academic career up until pretty close to the time that I that I ended up applying for a job elsewhere. I had always really thought of myself as a career academic. But these days, even if you have a research program that you have a lot of confidence in, and even if you have great support as I did, from your advisor and your your committee and other people around you, still almost nobody is really sure of an academic job just because it's a fixed supply market. So I always felt like it was good to be developing my other interests, to be thinking about how they could turn into careers. A lot of those interests did revolve around tech. I was sort of a hobbyist programmer since I was a teenager, really, and had a lot of interest in the way that we try and put models that do some kind of reasoning into these sophisticated, what used to be called expert systems. Now they're sort of all called AI because that term paints with a really broad brush. But that's something that I always sort of been interested in, and I had kind of gotten some opportunities to bring some aspects of those things together because there is a lot of computer modeling that happens in, particularly in the in the Logic and Philosophy Science Department at UCI.
Dr. Nathan Fulton: And so I got to do coursework with just absolutely outstanding faculty. And I also had some peers who were in the program at the same time that I was that taught me a lot and were really generous with kind of helping me get my footing with designing models that would show the behavior of agents. So I had these entrepreneurial ideas like app ideas. I was sort of taking a look at the old market for tech, and there's really two sides to that. One side of it is in order to try and provide myself with options because in academia, but also in any job market where people really love what they do, that's where it's seen as a really great and fulfilling kind of work to do. But that also pays well and includes some really positive work environments. Those jobs are always going to be really sought after. So if you're someone who spend a lot of time reading the Stoics, you know, and you want to cultivate a good life for yourself, and you understand that there are going to be some things you're in control of and some things you're not in control of. And then you've also done a whole bunch of game theory and rational decision theory. One way to kind of protect yourself is to explore several different avenues, because even if none of them are very high percentage, if you're exploring three or four different 15% possibilities, you can get your collective possibility of a strong outcome of over 50%. So sort of on the lookout for a few different things in order to try and give myself those chances to get lucky and land in that optimal position.
Dr. Nathan Fulton: But another thing that I started to get concerned with was just that, supposing that I did secure a position in academia that was potentially a tenure line, that really had a future doing the stuff that I loved, I would be taking that position from someone else. For a lot of reasons that are just sort of, you know, luck of what my interests were and certain industries that have emerged as providing a lot of economic opportunity right now, I had this option to have a number of different avenues so that one of them is going to be able to get a little bit lucky with and find a place to get something started. And some of my colleagues who were and continue to be just incredible scholars didn't necessarily have quite as big a range of options. So they were going to need to get lucky within this somewhat narrower range of scenarios. And so I just wasn't totally comfortable competing as hard in those slots because it was a little bit of a catch 22. If I did get one of these coveted spots, then I would feel like I was taking it away from someone else because it's a little bit of a zero sum game. And that just wasn't what I preferred to do, given that there were some other options.
Kino: I really like what you said about pursuing different opportunities that might individually appear improbable, but collectively would result in a higher chance of a good outcome. As I'm preparing this podcast and talking to people about it, one common thing I hear is that people would say, You know, I am in a career now, but the way I got here is through some particular niche path that some some specific networking that just happened to work and it's not really replicable. And so I don't know if I'm a good person to be giving advice. And my response is that this is probably the norm. Now for humanities PhDs pursuing outside careers because there is no tested and true path to follow is there is no set of set of procedures that if you just follow this each step, you'll be fine. Everyone got to where they are now through some low chance path. But if we aggregate all these parts together, we can create a substantive body of opportunities. So I really agree with the attitude that you were talking about earlier.
Kino: So as you're looking into our tech careers near the end of your graduate school, did you talk to people around you about it? Were they generally supportive or was there any campus resource that was helpful?
Dr. Nathan Fulton: I didn't actually use any formal campus resources or anything like that. It's very possible that they were available and I wasn't aware of them because I was still a lot of my energy was still just sort of focused on my my dissertation and my teaching. I was having a lot of conversations about careers with people. They were they were very informal. And a lot of it was just the the General Department gossip that you have where you you hear about people who are teaching, but maybe they're teaching at a private high school that has that actually has philosophy classes, and then they additionally teach some English classes or maybe history classes or something like that. Certainly I knew some people who had already gone into tech from my department, and that was reassuring, just basically to have evidence that you don't have to have an A graduate degree in software engineering to land these entry level, but with with good prospects and jobs in tech, because with what we do, the level of qualification is so high. You have to have a doctorate, you have to have publications that's very competitive. Ideally, you should have a strong teaching portfolio and have been the instructor of record. So you should already have practical experience doing what is what a lot of your actual time and energy in the job is going to be taken up. By doing so, it would be easy to think, Oh, it must be the same for other kinds of jobs. You should have the high tier qualification; you should actually already have some experience. Maybe it's an internship there or whatever. But one thing that's important to understand is that with a lot of areas of technology services, the demand radically outstrips the supply. And there's lots of different variations where there are other skills that you can bring in, in addition to just basic competency programming in one or two languages that are familiar to whoever you're interviewing with being able to demonstrate communication skills and other skills that are difficult to formally test but that will become pretty evident in interviews, that sort of thing. And then analytic skills and particularly being able to demonstrate analytic skills on the fly, those things can sort of supplement even a pretty basic facility with programming and understanding databases and understanding networking and other things about computers. So it's just reassuring to hear that people I knew whose experience with computers was was similar to mine, at least in the same ballpark. We're being successful there. I think that was the biggest sort of thing.
Kino: Were there any jobs that you seriously considered and then decided against?
Dr. Nathan Fulton: So I had jobs at times when I didn't have teaching appointment or I needed to have some other source of income. I had had jobs doing some of the kinds of things, service industry things obviously, waiting tables or barbacking, but also something that I think a lot of graduate students do, which is test prep tutoring or other kinds of tutoring. And I pursued some opportunities with a couple of different companies for different kinds of school subject tutoring for high schoolers or that sort of thing. The thing that I did the most was working for Kaplan, a large test prep company doing test prep classes for the SATs, but also for the LSATs and for other kind of subject specific standardized tests of various kinds. But that sort of thing was a little bit of persistence, and especially if you are willing to commute, those jobs are available, but they're not really career building jobs in most cases. There's maybe opportunities for a few people to move up in the organization and transition their management roles, but they really are almost kind of intended to be these more short term jobs. And then in terms of some of the kinds of jobs that I mentioned a moment ago, I was seeing people get and that I thought about pursuing in tech, I actually wasn't really thinking of myself as on the job market for like a full time career building job.
Dr. Nathan Fulton: At the time that I applied to Cycorp, what would be my first tech job, I was still very much in the middle of my dissertation work. I think it was another year and a half after I started working at Cycorp that I actually defended. So I took that job before I was done. So I wasn't quite so, I hadn't even gotten to the cycle in my academic career where I was thinking, okay, I'm going to defend in the spring, so this fall I'm going to apply for jobs for the following fall or anything like that. I wasn't even really in that place with academic jobs. For that sort of stuff, I wasn't really looking. I just knew that outside of post-secondary education, you know, outside of college level education, teaching jobs were not appealing, the test prep jobs, the tutoring job. People that I knew who taught public school, which just seemed exhausting and a lot of the time even tough to make a living at. So I was wanting to move away from education generally, not just higher education, but I wasn't quite yet looking at other, like you alluded to earlier. A lot of it is kind of like a very niche networking sort of story.
Ontologist at Cycorp
Kino: Your first job after the PhD was at Cycorp. I know that they are one of the very few places that advertise on PhilJobs, which is a job advertisement website for philosophers that's primarily for academic positions. Is that how you heard of Cycorp?
Dr. Nathan Fulton: That's actually not directly how I found it. It might be indirectly how I found it in that I had a close friend from graduate school, actually, someone from LPS who had started working there. I was working on my dissertation and my friend just sent me a bunch of messages about the place: This place is really remarkable. This is not really what I expected from a programming job. There's a lot of highly conceptual discussion behind everything that we do. They definitely hire philosophers, but also you're a LISP packer, which is to say I had some familiarity with a historically very important to AI but now slightly obscure programming language called LISP and Cycorp is, or at least at the time I worked for them still was, one of the few fully commercial entities that relies very heavily on LISP as an internal dialect list that they use for their programming infrastructure. So that combined with the fact that I'd spent a lot of time teaching first-order predicate calculus, made me a good fit for them. In order to prepare for [the interview]. So I will say that Cycorp is not characteristic of companies, even artificial intelligence companies, in terms of what they're really looking for. So I had some good background, because I had a personal connection, on what to prepare for, which involved some slightly abstract and highly technical, in a formal way, computer science concerns around concepts like recursion and time complexity. They were looking for people who were prepared to talk about some fairly intensely technical aspects of how the algorithms they were writing for these things worked. So that was sort of on the programming side, a lot of what I was thinking about and trying to prepare for.
Dr. Nathan Fulton: And then on the philosophy side, sort of a combination of things with logic and set theory at the core, basically being able to switch really smoothly back and forth between thinking about a problem in logic from the theorem proving perspective of predicate calculus, but then understanding what that would mean if you represent it in more of a set theoretic way and being able to transition between those really smoothly. And then also think about that in terms of higher-order logic, particularly second-order logic and what you might call meta-ontology, categories of categories, and understanding the relationships, but also understanding primitive relations and topics in logic like advanced subsumption and subsumption that works via predicates in addition to subsumption that works via classes and a whole bunch of topics like that. And so the Cycorp interview, I doubt there's another interview like it for almost any organization anywhere. And probably if there are other interviews like it, they are for postdocs in academic environments more than they are for commercial enterprises. But one way or another, thinking about all that stuff seems to have been enough and having a strong recommendation from someone who was already working there. But I will say that I know that, although I didn't happen to, a lot of the people who work at Cycorp do end up applying through those PhilJobs postings.
Kino: You were an ontologist at Cycorp. Can you tell me what that is? What would a typical day of an ontologist be like?
Dr. Nathan Fulton: A typical day at Cycorp is doing one of about three, I would say different kinds of activities. What you probably spend the most time doing is something that is effectively data entry, but it is data entry that you have to be really good at logic and at translating natural language into logic to do, to the point where they primarily hire philosophy PhDs, because that's just the most reliable way to find people who can do that. There are, of course, people who are coming from more of a mathematics background who can correctly conduct sophisticated theorem proving with logic, but the clear translation that's capturing what sentences in natural language you're trying to represent and capturing certain kinds of nuances is, that's not something that people coming from mathematics are necessarily going to have as much training in. And particularly when you get into things like understanding the subjunctive in English and understanding whether you need to think of it modally or whether there's another formalism that is helpful, deontic logic, things like that. It's pretty hard to find people who are really comfortable doing that, and will just be able to [do it], day in and day out, for hours at a time, outside of philosophy PhDs or once in a while a very gifted [person] who has an undergraduate degree in philosophy from a really strong department. So yeah. So you're basically you're taking some data that's been gathered just by people doing any of the different things that cause data to be gathered, filling out forms, displaying some kinds of behaviors or whatever. And you are translating it into this extremely powerful, expressive, logical language called Cyc-L. That's what Cycorp uses internally. But it may just be a large set of pretty mundane data. And so you may spend basically days on end just sort of doing this really hard, but also very repetitive data entry.
Dr. Nathan Fulton: And then of course, there're the periods in between that where you're sort of making it possible for that to happen. You are creating new predicates, new classes, whatever, other logical architecture you're going to need to input the next data set if it's something new and also possibly writing rules, they're basically predicate calculus, rules of inference that will allow Cyc's inference engine to do theorem proving and derive new formula, expressing further facts that were not in the initial data set. So yeah. So you're turning data into facts by essentially re-expressing it in a semantically richer notation. And then that notation is designed to enable an automated system to actually generate further facts that are strictly provable, they're deductions essentially. So those are sort of the first two to pieces.
Dr. Nathan Fulton: And then the third piece is helping to figure out how this process of inputting this data, semantically enriching it so that it's sort of more fully representing features the world that's supposed to be capturing, enabling the system to derive new facts. None of that is useful unless you can figure out how it becomes some kind of a deliverable or a product for whatever client you're trying to help out, discover new things from their data, maybe even just help them be able to query their data more efficiently by putting it into a format where you have a much more flexible ability to ask questions that are just based on the data you already have, whatever it is. Just as you need basically any people with PhDs to enter new assertions into the system, it's also really, really difficult for people who don't have philosophy PhDs and several weeks of experience with Cyc to ask it anything, to write the queries. And that doesn't actually make for something that provides value to clients. So you have to kind of put that logical querying layer in the background, behind some kind of an interface that simplifies down what are the values for variables that we need people to fill out into some little boxes and maybe they're autocomplete dropdowns and, and then once they plug those in and we have values for a few variables where it's really this particular system that we're building for someone is really just asking the same query from a formal perspective over and over.
Dr. Nathan Fulton: So that's how you can kind of turn Cyc into a product for a specific application. So that's the three things that I would say, generally speaking, an ontological engineer or ontologist at Cycorp spends most of their time doing. There's also inference programmers who, as you might imagine, spend their time [doing] two things: improving the inference engine generally, making sure that all the logical patterns Cycorp really needs to be able to perform inferences are there, that they're accurate, that they are fast enough, that the system performs quickly enough. And then the other side of it is, to a greater or lesser degree, depending on how involved you are with with user experiences and that sort of thing, working on these ways of figuring out how can we have some kind of simplified way of plugging some information in and tell people who don't have all of this background and comfort with these highly complex formal patterns of reasoning, how can we ask them for just the right pieces of information and then right a little bit of bridge code that turns those pieces of information into query and a sort of very, very sophisticated theorem proving system.
Semantic Architect at Indeed
Kino: You now work at Indeed as a semantic architect. Can you tell me a bit more about what that's like?
Dr. Nathan Fulton: Yeah. So I think probably a lot of your listeners are aware, but for for those who aren't, I'll mention it. Indeed is a company that started out doing job search, providing you with an interface to search for jobs and now offers a wide range of products that are, to a greater or lesser degree, adjacent to that. So we still, of course, provide job search and that's a flagship product, but we also provide more tailored experiences for people who are in specific occupations, where we know that you're looking for a job as a nurse and we can actually show you a list of certifications and ask you which ones you have, because our system is really informed about what certifications are required by nurses. And then what's your medical specialty, what kind of work environment do you like, what kind of benefits you're looking for, all those sorts of things. So we have a lot of different products that try and go beyond search to provide you with a really efficient experience that shows we understand what you do and we want to help you find your dream job. A lot of our goals right now have to do with wanting to reduce the number of steps that the job seekers or employers have to go through and really just help them find each other as quickly and precisely as possible.
Dr. Nathan Fulton: So what I do is so I work on the taxonomy team for Indeed. So the taxonomy team provides these couple of different taxonomies of things that are important to know about a job and about a job seeker in order to match them effectively. We have a taxonomy of occupations, so all the different kinds of jobs and it's a true taxonomy in the sense that it's somewhat, it's not a strict tree or being a directed graph technically arborescence. It's not strictly that. There are some cases where a particular occupation might have a couple of more general occupations. So, for example, a law librarian is both a legal professional and a librarian. So it's not strictly tree-like in that it just branches down to more and more specific jobs, but it's fairly tree-like and it provides a basic way of categorizing things. But then there are many, many other things that we track. The qualifications that people have, the benefits that they want.
Dr. Nathan Fulton: The thing about job descriptions, which is and this is something that I think you're going to find in any large natural language corpus, even if it has a common theme and some general elements, is that the prevalence of features is not a reliable guide to their importance. For example, aircraft mechanics require a license that's called an Airframe and Powerplant License. Every aircraft mechanic has to have this, in the United States. You can't work on an aircraft without it. [But] you absolutely cannot assume that 100% of job descriptions will say "must have your Airframe and Powerplant License". That doesn't mean that those jobs don't require it. They definitely, absolutely do. But because they definitely, absolutely do, a lot of employers feel it's too obvious. There's no reason to mention it. They don't have to mention what kind of license you need even though to you and I, Airframe and Powerplant License, like some of the technical terms there, we can guess at what they mean but it's a little bit obscure, we certainly wouldn't have guessed that that was what the license was called. But every mechanic, aircraft mechanic knows that. So they don't give that. We have to know. Our system has to know that every time a job in this occupation comes up, it requires this license. And we should not show that job to anyone who doesn't have that license. And then we want to be able to achieve the same level of understanding as people's resumes or profiles that they make on Indeed, where we know what all the licenses that they have are, what their educational qualifications are, things like that. We know as much about them as possible. We want to capture that. Even in cases where people leave things out. Resumes in a certain sense are often more complete in that people are a little bit more thorough about listing their qualifications than employers necessarily are with job descriptions. But on the other hand, some people are very thorough and they list a number of jobs that they've had. And if they've changed careers in any way or if some things are the jobs that they would really prefer to have and others or things that we're sort of filling in for, for a period, then it can be really tough for us to disambiguate that and know what's important.
Dr. Nathan Fulton: I work on solving these problems. My department figures out what are the categories of jobs, what are the features within jobs that are really important? And then I provide the basis for inferring from whatever we can gather, whatever people do say that we can pick up and directly extract through natural language analysis, what further things can we deductively infer based on or our knowledge graph of connections between a given occupation and the license it requires, and all the certifications that are required maybe for particular specializations within it so they're likely to be required, but not as certain to be required as the licenses. What level of education is preferred by employers? What level of education is probably the absolute lowest where you're going to be successful in applying to these jobs? Things like that. And then it starts to get more complicated beyond that because it may mean that you need to not only identify what license someone needs, but what state do they need to have that license in. And so much of this is completely common sense to humans. You know, you're going to say, well, the job's posted in Texas, so they probably need a Texas driver's license. But the system doesn't know that. Computers don't know anything that you don't tell them. And we have to be able to process this stuff in an automated way because Indeed is the world's largest company for this kind of thing. The number of jobs that appear and disappear on a given day is, you know, on the on the order of millions and the number of jobs that Indeed is tracking in total on a given day is in the tens of millions. So as these jobs come through our system, we have to be able to say, okay, we know that this is an abbreviation for this license, even though it just these three letters and it doesn't include the word license. Nonetheless, it's an alternative label for this thing. We can identify that. That's what the taxonomy department that I'm embedded within does. And knowing that this job requires this license, we know that it requires these skills because in the course of getting that license, you have to demonstrate these skills, so we can add those skills in. And now all of a sudden, we have a richer picture of the job than we did before. The job didn't mention any educational credential, but we know that for this occupation you have to have a bachelor's degree, that you couldn't actually have gotten that license without it. So to pick up on that clue.
Dr. Nathan Fulton: I designed the relations between things like occupations and skills or credentials. And I work with the engineers who build that inferential layer and to try and get as much information. So there's the stuff that is explicitly stated that we can just pull out of the job descriptions. There's the stuff that is obvious to almost anybody, that an adult human probably can figure this out. And then there's the stuff that's a little bit more professional insider knowledge, to really clear to the people who are qualified to be applying for the jobs. It's really clear to the employers, but we need to make sure the system understands it as well.
Kino: Do you need a lot of programming skills?
Dr. Nathan Fulton: Well, yes and no. I think someone could do the key aspects of my job without writing any code. But the way that I do my job. So first of all, the way that I do my job doesn't tend to involve writing code because I use scripts just to automate tasks, maybe to write a demo to sort of say to an engineer, here's a Python version of what we need this algorithm to do. You can see that we're putting this stuff in, here's the stuff we want to get out. Here is how you can use some of the connections in our ontology to get from A to B. And so then they're going to integrate that with our system's libraries and our system's native programming languages. We're giving them just a sort of Python example is really helpful. And in general, being familiar enough with programming concepts and also data structure concepts, which are not quite the same but are kind of closely aligned, is just really, really helpful for being able to communicate quickly and effectively with engineers. But it's not a software engineering job. And the programming I do is typically either examples like I just talked about, or maybe there are times when I want to upload some data and there's a code-based way of doing that more quickly. And then there are tasks that are, where you're helping engineers now in order for your colleagues to have to do less programming or deal less directly with data. We try and build tools where it's some engineering up front, but the goal is that then we're going to have a lot of just autocompleting menus and be able to sort of take one to unplug the data that it outputs into another tool sort of seamlessly. Because we do have, we have a lot of amazing team members who make huge contributions, but they don't program. They don't have that particular technical background.
Learning More About This Path
Kino: So suppose someone listening to the podcast finds what we just talked about to be super interesting, but it's their first time hearing about anything like this. How do you suggest they start learning more about this career path and eventually hopefully moving in this direction?
Dr. Nathan Fulton: Well, if they were at an institution that has either a Computer Science Department or an Information School, so where you get an M.LS., M.IS., Master of Library and Information Sciences. If they're at an institution that offers strong coursework in semantic data, linked data is another sort of area, things like information management. I know they sound like kind of vague terms, but they actually do, if you're looking in the right course catalog, then these are good keywords for this kind of thing. And you take a look at them and they either don't have a lot of prerequisites, or you can maybe take one introductory programming class, which by the way, is also always helpful, and then meet the prerequisites to do those courses. That'll just introduce you to a lot of foundational technologies and methodologies that will help you spot jobs doing this kind of thing and help you just be oriented for interviews so that you can then talk about how other skills that you have. Analytical skills, communication skills are going to be really helpful for this.
Dr. Nathan Fulton: If you are in philosophy departments, don't dodge that assignment teaching Intro to Logic, partially because if you are going to do semantic data management, your logic skills need to be fresh. You need to recognize the kinds of formal mistakes that people are likely to make when they take your data and implement it and turn it into user experiences you need to spot if they are affirming the consequent or doing any of these kind of like boilerplate things that we're all really prone to until we get some practice. So just make sure your skills are fresh. But also it'll just be so helpful for you down the line with articulating to project managers on other teams how the very sophisticated structured data that you are providing works, what the value it's adding is, and how they can get the most out of it. It's going to be a lot like explaining to undergraduates how to translate a sentence into first-order predicate calculus or back out. So that's just really good skills building. There are, unfortunately at this time, not a lot of books for semantic data or ontology.
Dr. Nathan Fulton: I would say, as a career kind of overview for getting into semantic data. Heather Hedden's book, The Accidental Taxonomist is still sort of the standard, and that can also be a really useful book because entry-level jobs in this often are taxonomy jobs. And then you can leverage your logic and analytical skills to move to the more design-oriented jobs that are ontologist, knowledge manager, semantic architect. There's a bunch of different titles, so unfortunately there's not one search term that will really nail it. Another couple of books that are very good are Hendler and Allemang's book, Semantic Web for the Working Ontologists. It's a little bit more technical. It will be helpful if you are comfortable with basic data structures, even if you have done stuff like build web pages as a hobbyist, just sort of that level of comfort with computers and specifying data is about what you need. But that's a really good book, especially if you prepared to make the investment for the third edition or you can find it in your school library. Also, a more recent- well, the third edition of that book is quite recent. Also quite recent is Panos Alexopoulos's really excellent book, which is I think is called Semantic Modeling for Data. That's a great best practices book to read if you're pursuing a job where you're going to be in charge of semantic data. But also reading through that book is going to tell you, these are the things that you're going to need to be policing people on. These are the best practices you're going to have to fight for. So can you be a stickler about this? Can you be the person who says to a manager, higher than you in the food chain who really wants the data set now, no, we can't send you this controlled vocabulary for you to include until we have written the definitions for every term. Because if we don't do that there there are going to be negative consequences down the line. Is having those kinds of conversations over and over again something that you're up for? So that's another one, good one, just to get a sense of that.
Dr. Nathan Fulton: And then one more thing that I will say about it is, if you think that you would be interested in working in finance or working in biotech, especially pharmaceuticals, then you have a broader field to investigate, especially if it wouldn't bother you to do something that is technical but kind of repetitive, a lot, as long as you're getting well paid for it, then there's really good career prospects there. But, you know, a lot of people in philosophy are not necessarily comfortable working for large financial institutions for lots of reasons. Philosophers of science may have some reservations about working for pharmaceutical companies that may just be purely epistemological. They may just have to do with epistemic terms. So, you know, a lot of it is also about who do you want to work for? I will say, Indeed is a great company to work for. Their company culture really is built around the idea that we help people find jobs. The day to day company culture is very inclusive. It's very positive. All of the people that I work with are very smart, but also very kind and I love it here.
Dr. Nathan Fulton: I know there are a lot of other great workplaces out there. Look out for that as well. Be aware of the space that you are trying to move into. In addition to building up all of these skills, research the space that you're moving into. If there are things about the working environment that you don't think are great, but you think that you could put up with them, ask for more money. Because probably other people don't love them [either]. So being willing to put up with a certain cadence of deadlines and a certain repetitiveness of the task or whatever that is, that is worth money. And if you find that something seems like kind of a dream job, then think about managing your expectations and being willing to, of course you've got to make a living, but being willing to take a little bit less than you might have otherwise be been seeking in order to get your foot in the door. Because if it's a job that you can stick with and build up that experience. Then in tech, in finance, in biotech or health care, there are going to be advancement opportunities. Your first job doesn't have to do everything for you. It can be here. It can be your foot in the door.
General Advice on Alt-Ac
Kino: What other advice do you have for graduate students or fresh PhDs in the humanities who are thinking about pursuing our tech careers?
Dr. Nathan Fulton: Don't undervalue your graduate degree. You need to get your foot in the door. In an industry you initially, when you're looking for your first job out of college, it may not pay that well. It may turn out to not be as intellectually stimulating as you hoped once you really get into it. You still need it because they need to see that you can get to understand their industry or their workflows, the sorts of things that you do day in and day out. Once you can demonstrate some experience and it can be as little as a year, especially in fast moving areas like tech, that experience that will get you into your next interview and let you start moving up, will let you ask for a little more. And your graduate degree is going to be a multiplier on that. Master's? Great. That is more money. PhD? Great great. That's rapid career advancement. So initially you have to understand that they don't know if you can do this thing that they do. Once you show you can do this thing, then you're going in for interviews and they're looking at your resume and seeing you can do the thing they need you to do, and also this incredibly hard, grueling thing which is write a Doctoral dissertation. And that just means that you're a potential for getting more responsibility, which comes with more compensation but also comes with more professional freedom, more possibilities of advancement that is really meaningful. So don't worry too much about advancing your career to what you need to be with the first job that you take. Because your graduate education will mean that your career will advance more rapidly, assuming, of course, you can learn all that officy stuff -- being able to communicate well with people, stand firm on things where, you know, it's going to be costly to do something other than the right thing, but give ground on things that would be nice to have but aren't necessary. Like all of that basic being in a work environment thing. Organized and either being able to meet a deadline or clearly articulate why you can't meet the deadline, that sort of stuff. You need to do all that stuff. But if you can, if you can do all that stuff, which you probably can if you have been able to teach a course, write a paper, in any discipline at the graduate level, then you're going to be fine. And once you get going, you'll move up fast.
Kino: Thank you, Dr. Fulton, for sharing your wisdom with us today.
Dr. Nathan Fulton: Thank you. I think that this podcast is a really, really great project and I'm excited to see where it goes.
Kino: Thank you for listening to our second episode with Dr. Nathan Fulton. I have been your host, Kino. You can find me on Twitter at off campus FDs or email me at off campus pot at gmail.com. I'm always looking for suggestions of guests to invite questions to ask and topics to explore. I'll see you next time.