Alex Rutherford is a freelance data scientist and entrepreneur with a PhD in Physics from the University College London. He has subsequently undertaken post-doctoral work in complexity science and computational social science using computational techniques to understand why ethnic violence breaks out, how large groups of people can work together remotely and how constitutional reform takes place. His work has been published in Proceedings of National Academy of Sciences and has been covered in the New York Times and Nature among others.
Alex worked as a data scientist for the United Nations in New York for several years applying computational techniques such as natural language processing and network analysis to aid and inform the development and humanitarian work of UN agencies and NGOs. This has included field work in Mexico, Jordan and Brazil, collaborations with numerous blue chip companies, presence at high level UN events and a handshake from Ban Ki Moon. Alex has lived and studied in Coventry, London, Damascus, Boston, Dubai, New York and Silicon Valley and speaks passable Spanish and Arabic.
More recently, Alex is the founder of Data Apparel, an organisation selling custom, ethical clothing that uses the power of data and visualisation to promote empathy and debate among global citizens. He is an active Twitter user and blogger. More information and contact details can be found at alexrutherford.org.
The problem that I eventually settled on [for my PhD] fit a few different criteria that were important for me: 1) something involving computers; 2) something that had some real world applicability; 3) be in a big city with the opportunity to stay grounded.
– Dr Alex Rutherford, data scientist and founder at DataApparel
#PhDCareerStories #Story #Entrepreneurship
Hello, my name is Alex Rutherford, I am currently a freelance data scientist and entrepreneur.
I am very excited to be here speaking in the PhD Career podcast. I think it’s a wonderful resource and something I wish I had access to when I was a PhD student.
I’m going to talk a little bit about how I got to where I am now; my decision to start a PhD and my experience completing that PhD and the work that I did subsequently including data science work at the United Nations.
I was always a very inquisitive child. I wanted to understand how the things worked and take them apart. Luckily, my parents were very patient so when i decided to take apart an antique grandfather clock for example they managed to indulge me. And they also left around books by Stephen Hawking and the Space Encyclopedia and that was something that really caught my imagination and made me want to go and study Physics at university and learn about stars, black holes, quantum mechanics, relativity and all these amazing ideas that I had read about.
I went off at the age of 18 to the University of Warwick in the midlands in the UK and pretty soon I found that quantum mechanics, relativity and cosmology are really, really difficult and designed for people that are much more intelligent than I am. But something happened towards the end of my first year that put my life on a different track. I remember it very well. They asked all of us to come in for an assignment in the computer room. It was a banquet of old windows and tea machines and they said: “Ok, we are going to simulate the path of a projectile so imagine throwing a ball in the air. We are gonna solve those equations, Newton’s equations of motion, using a computer. I’m going to write a program in the QBasic programming language, an ancient programming language. And I’ve never done anything like this before and I struggled, but I just loved that idea of taking these instructions, breaking them down and giving it to a computer and having that computer combs through that algorithm and give you what you want at the end of it. More or less at that point, I decided that I was not gonna be someone who worked in a lab or derived complicated string theory equations on the board but instead I knew I was gonna use computers in some way.
So I made the decision after I graduated that I wanted to go and probably do a PhD, but it wasn’t something I wanted to rush into after all that study. I was interested in how we could use these ideas from physics, writing down equations for systems and how they work using some physical motivation, thinking about processes like diffusion, random walks… The only place that was being applied at the time was in finance. There were a lot of pioneering physicists like Benoît Mandelbrot and Emanuel Derman who were very successful in using these ideas of random walks and random processes to model predator relatives and that was something very fashionable at the time. Rather than jumping into a PhD I thought let’s just try and start working and see what it’s like in the professional world. So I worked for a short while as a stockbroker and I very quickly realized that I like academia and its daily intellectual stimulation. I decided to apply to a few PhD programmes and settled on University College London (UCL).
The problem that I eventually settled on fit a few different criteria that were important for me:
1) something involving computers. I wanted to learn how we can use parallel programming to distribute the programming to solve problems. At that point it was important to me that ‘the bigger the better’. The bigger the problem, the bigger the system to be simulated, the more processes we have crunching away the better.
2) something that had some real world applicability. Something that was likely to be picked up and used instead of something to be left as some intellectual oddity.
3) be in a big city with the opportunity to stay grounded and not to be in an ivory tower.
I was very lucky to find an opportunity working on radiation damage in nuclear fusion reactors. The idea here is that you have a nuclear fusion reaction which is basically like a star which you somehow need to contain in a lab. This throws up all kinds of challenges. One of them is: how can you design something in the wall that can absorb all types of radiation coming out of it.
My task was basically to simulate millions and millions of atoms individually, write down Newton’s equations of motion that each one - after they get smashed in to and transfer their energy and simulate that thing nanosecond by nanosecond or rather femtosecond by femtosecond even on a big bank of computers I’m trying to understand that process.
My PhD experience is probably like a lot of people’s experiences, something that I found tough at times. There were definitely times when I wanted to give up, where I questioned whether I belonged there. It is a hard process and I don’t think there is any way to get around that. But I have absolutely no regrets. It was a wonderful time to learn to learn, to be around smart people and the thing I’m most grateful for is thinking what I wanted at the end of this process.
I took a year’s break between my graduate and my post-graduate and that gave me a good sense of perspective on how PhD skills can be useful, what I can expect from working a job nine to five and it gave me a chance to travel as well and broaden my horizon a bit and remember that there is a big world out there – outside of academia.
I had a very clear idea that I didn’t want to work in academia for the rest of my life. But I wanted to find a way that those ideas of physics, modeling and computer programming could be applied in a useful way that has some social value. And the other piece that I had very firmly in my mind was that I wanted to make use of all the things that come with being in a large and established institution; a good university in a big city where there are actually a lot of opportunities for you.
For example I took language classes at the language center both in Arabic and Spanish. This was very nice for a few reasons: First I found it very hard to turn off my brain if I’ve been studying and reading and working through a lot of literature review all day. It was impossible for me to just switch off and sit in front of the TV and do nothing. So rather than obsessing and working all night, the thing that was working for me was keep using my brain but use it on something very different; use the other side of my brain and learn a language and practice that. The other piece is, UCL has a very good programme of graduate education that is compulsory whereby PhD students have to get a certain number of credits in transferable skills. A lot of people resented this programme because they saw it as a waste of time. But for me it was super valuable. I was able to take workshops on developing your CV, developing your communication and project management, the language classes also counted towards these credits as well as volunteering. I sat on the board of a charity in London that provides services to Latin American migrants. That was something i was inspired to do after traveling in South America before I started my programme. It’s a wonderful experience to see this other side of a big city and get some perspective on people outside of your demographic. The other thing that strikes you about the PhD lifestyle is that it is somewhat isolated and unique. You’ll speak to a very thin slice of society. People who are extremely intelligent, probably had access to a lot of opportunities, have spent a good few of their years thinking as a scientist, as a social scientist or whatever your department and your discipline may be. It was nice to have that kind of ongoing perspective and not lose touch.
The final piece that many people may relate to when you do your PhD is both a good and a bad thing, it cuts both ways. You have all of your friends who decided to go on and start their careers and not to do further education. You see them progressing and getting promotions and being well paid; very widely being better paid than you as a PhD student and that can be something quite disheartening. It takes some resolve to remember that you are doing something that you love. But there are some advantages that allow you to understand how that extra training, that higher education, that post-graduate PhD skills base - how that can affect in compared to people to don’t have it. To give you a sense of what are the values that I can add, what are my learnings here, what skills are gonna be useful. And that’s really the trick of being a PhD student and being a PhD graduate and entering that job market.
All of a sudden you are in a small minority of people who have that qualification. You have to put in some effort to sell those skills, to sell this qualification. A person who is looking at your qualification, when you eventually apply for a job, they may not be a PhD graduate themselves. You have to explain to them very clearly what that PhD skill adds and what you can offer them.
Typically that is very advances literature review skills, looking at how to research a very broad nebulous area and how to precisely summarize it and explain it to people; that is something that’s very common across all disciplines. And then you may have some particular technical skills. You may learn how to programme in a certain language, you may pick up a foreign language, you may have a particular knowledge of a sector, whatever it may be it is on you as a PhD graduate to understand what it is that you can bring to the table and to be able to describe and sell that.
I may make it up like I had a very clear idea of how best to approach a PhD and all this level headed advice. And that’s all very easy when you have this retrospective view but of course like everyone I struggled to find a good routine to stop myself from going crazy and getting overwhelmed. One particular piece that was hard for me was knowing when to stop working. You have this very long period, in the UK you have around three or four yours, in other countries it may be as much as five years, you have this big period of time and this very hazy goal at the end of writing a thesis and graduating and at some point for me temptation was to think how I could speed that process up. Again me coming back to university and after finishing up my first degree, it was if you had a lot of work at your desk you’d stay and just do it. But that doesn’t apply to something that’s so cerebral, something that takes so much dedicated thought. It’s not really possible to just sit down and work really hard and catch up in many cases. You are reading a lot of background material which takes a lot of time to sink in, you have to imbibe the huge body of literature that you have to take in and put it to your literature review that has been produced by lots of clever people in the past. That is not something that can be done overnight and it’s rarely something that can be rushed. One important aspect is a sense of patience as well. You may feel that you are not getting anywhere and success is very slow and achievements sort of come very sporadically. But that’s really ok. One of the things that helped me here was to take myself out of that context, whenever I could. I took the opportunity to travel as much as possible, I was lucky to live near Europe so i could take short breaks there and that also gave me a chance to practice my language skills. But I also became very interested in current affairs during that time and traveled to South America again, to the Middle East and elsewhere whenever I had the opportunity. More and more i began to have an interested of living abroad, current affairs and dovetailing with this volunteer work I was doing.
As I came towards the end of my PhD i wanted to find a way to melt those two together. It wasn’t very clear how to do that, but luckily the decision was made for me. As I was finishing up my theses back in September 2008 I landed myself a very plunge job doing financial consulting in the city of London. In the middle of all that Lehman Brothers collapsed and sent all of these shock waves through the financial system and I found myself without a job offer suddenly. Noone was hiring anyone and everyone was on edge. Just weeks before I was to hand in my work of art I found myself without the job I had banked on. Looking back this was really a great opportunity for me: I had finished my PhD, I had earned some time to develop myself outside of Physics and I made the decision to continue studying Arabic in the Middle East. So I packed a bag and spent twelve months living in Damascus in Syria - this was in 2008/2009, long before the civil war erupted. I spent wonderful twelve months there immersing myself in everything wonderful that city has to offer, learning about how to get around, how to navigate a new culture, learning lots of new history etc. One thing that was very nice there was not being surrounded by physicists. In my Arabic class tended to be diplomats or journalists or historians, people from very different backgrounds from myself; often from China, Malaysia, Jordan – all different parts of the world. It was a wonderful way to broaden my horizons. Then began the second part of my professional career which has been mostly spent living abroad.
After I returned from Syria I began looking for postdoc positions. Although I didn’t want to be an academic for the rest of my life I recognized that a postdoc position is actually a very good way to transition from one field to another. Very often I find people rush into their PhD topic and it may not be the path they really want to continue down. It may end up that something they enjoyed studying from the textbook when they were hitting the frontiers of new knowledge that it’s really not so enjoyable and the right thing for them.
It was much the same for me, I realized that I wanted to understand social systems and how we can model markets or political behaviour or crowd behaviour or the diffusion of languages, how history has changed using computational methods. On the other side I hadn’t done a degree with a lot of computational work so I had a lot of gaps to fill in that I wouldn't have had if I had been a computer scientist. With that in mind I found a unique institution called the New England Complex Systems Institute which is in Cambridge, Massachusetts and part of the fantastic academic ecosystem of MIT in Harvard, north of the Charles River in Boston. The work I did there was looking at how we can take ideas from Physics, particularly condensed matter, and how atoms bond together and form clusters. How can we take this very well understood idea from statistical physics and apply it to people. An amazing ambitious piece of work my predecessors there were able to predict outbreaks of ethnic violence in the former Yugoslavia using exactly this idea. We have this very general idea of how a system of objects groups together, reaches some critical behaviour. If any physicist is listening they will understand that this is a phase transition driven by an autopractor. They took the same idea ad found that when you find a critical size of people you have an outbreak of ethnic violence. When a group is small people are very well mixed and there is no incentive to inciting violence because no one has a local majority. On the other hand when groups are huge people are actually separated and sort of ghettoized. The critical thing is somewhere in the middle in groups that are large enough to exert their own authority but not big enough that they can win and outnumber the other groups. And that’s when violence breaks out. You can test this system by running simulations and comparison to actual reports from newspapers on where violence broke out. It works with an accuracy of 90%. This kind of work was amazing to me because all of a sudden all these things that i had learned in dusty lecture halls that were developed by a German physicist in the previous century, all of a sudden this was being replaced with something really important and with consequences for history and humanity. My work during this time was to extent that by looking at Switzerland which is a country that has (almost) no violence and that is surprising because you have lots of different religions and language groups there and my job was to look at the effect of the natural boundaries formed by the terrain, the mountains, and also political boundaries which Switzerland has which is able to give autonomy to different linguistic regions.
The years of 2010/2011 was an exciting time in Cambridge because you have this amazingly high standard of science happening there. And what was happening at the time was what we know as Data Science now or as Big Data. People were beginning to extract data and information from people’s behaviour in innovative ways and doing sophisticated statistical analyses and modeling and analyzing natural language or formulating a network from that and doing some calculations on that network structure. More and more I began to develop these methods and going down the track of computer science. Another opportunity came up then which was to work at the Masdar Institute of Science and Technology – at the time a new venture being developed in Abu Dhabi, one of the United Arab Emirates (UAE) better known of which is Dubai in the Middle East. This institution was founded to help the knowledge economy in Abu Dhabi to be better prepared for the eventual end of the oil economy. It was set up in collaboration with MIT - an exciting new endeavour. I worked at the time with Professor Iyad Rahwan who is now working at MIT. It is actually a funny story how I came to be working for him: He was looking for a computer scientist and he wasn’t sure that physicists could do useful work. We have this joke that physicists like to always model everything as being like an atom. If you are thinking about a farm then you’ll say well let’s imagine we have n spherical cows with a certain radius. We have this reputation for making very simple assumptions about things and deriving models which at the end don’t have any use. So I had to work quite hard to convince him. He’s originally from Aleppo in Syria and I spoke to him in Arabic and I think that was really the thing that convinced him. He took pity in me and said: Okay, I’ll give you a chance.
My second postdoc appointment was working in Abu Dhabi on the diffusion of information on social networks. This was inspired by a challenge setup by DARPA, the big research funding agency in the US on the 50th anniversary of the establishment of the internet. The idea was: How can we see what the internet can achieve? Other intelligence gathering tasks that would otherwise be impossible. They set up a challenge with ten big red weather balloons and they placed them at random locations in the mainland US. To find the locations of those by a sort of individual team based in one place is described as being impossible by conventional methods; or at least doing it in any reasonable amount of time. It was an open challenge to use the internet to see how that could be done. My colleagues at MIT were able to - in the space of 48 hours - set up a site and a reporting mechanism and an incentive mechanism that they found all of those ten balloons in these random locations. The critical question was: is this something that can be repeated? Since this was a one off and we didn’t know if this was the case. Using a lot of data collected during that challenge and very high resolution population data I ran simulations to basically repeat that experiment 10.000 times and see how many of those 10.000 times it actually repeats. And this is the value simulation being able to ask ‘What if …’ and change the parameters and see the degree of success.
I offer to any PhD students who are looking at their options for after they graduate to consider postdoc appointments. If you really have enough of academia you are maybe tempted to just leave and never come back. But postdoc opportunities can offer a lot of advantages: By that point you have your research skills much better prepared than when you are a PhD student. You also have a lot more freedom and the ability to transition to a slightly different speciality. Quite often supervisors will give you the freedom to pick up these skills. You don’t necessarily have to be the perfect fit. Because of you have completed the PhD you demonstrated that you can teach yourself new things and work independently.
I was very happy to publish papers and to have time for self-directed learning and fill in the gaps and develop the skills that I felt I didn’t have. Still i felt strongly that data science was the missing piece of the puzzle for me. It was literally this new shiny thing were physicists were able to take all the things that they learned and understand data, find patterns and insights in there and provide value and make use of all those things that we’ve learned. It is such a hot topic now the classic Harvard Business School article that describes Data Scientist: The sexiest job of the 21st century. I can say having worked as a data scientist sexy might be a bit too strong of a word but it is extremely rewarding.
I worked for three years in New York for the United Nations. This might be not the most obvious place you can think of as hiring a data scientist - typically it’s the Spotifys, the Amazons or the Googles that come to mind. But very interestingly the kinds of work that’s done to understand customers and their preferences and behaviours that’s going on in the private sector, there is a big overlap with governments, development agencies and NGOs. We want consumers to be healthy, happy and prosperous so they spend more. But that’s the same job as a government or a development agency. The organisation I worked for is called United Nations Global Pulse which was founded with the aim of taking these ideas and data collected around people and in a safe and responsible way taking those ideas and applying them to development problems.
For example, can we look at what people are saying on social media and understand why they might be hesitant to vaccinate their children? The same way as if I’m selling something like a car or a phone and I listen into people’s conversations, their sentiments and their sense whether this is a good purchase. Can we apply the same thing to vaccines?
We were doing a lot of social media projects using Twitter and Facebook data and then applying techniques that I learned previously in my postdoc work, like natural language processing, network analysis and all of these kinds of things. It’s an exciting time to have all this scientific knowledge and begin to apply it and see where it can add value within the UN system.
I spent a wonderful few years working in New York looking at how we can understand the spread of epidemics through data collected passively from mobile phone usage; or traveling to Zaatari Camp in Jordan to advise UN agency on how they can use new data sources to understand the needs of refugees; or even as diverse as using postal data from the UN agency that collects information on post (Universal Postal Union) and using that to predict indicators of a country’s well-being.
I imagine that lots of people that are coming toward the end of a PhD, physicists, neuroscientists, applied mathematicians and so on would have thought about data science. It’s a wonderful new emerging niche for people who have these numerous degrees to apply themselves and it’s fascinating how that idea has grown for someone who can make sense of passively collected data. I see finances being the first to capitalize on it and people who work for large organisations like Walmart and Amazon and have big logistics problems being the next beneficiaries. From then on it has really exploded: people who do advertising or look at translation or these myriad of other problems. On the one hand it is very exciting that we have a lot to offer as a community. On the other hand it is not a straightforward area to be employed in.
The first issue is that data science is not very well defined. Data scientists can have a lot of different backgrounds and can do very different kinds of work. A job title that has the description of data scientist requires you to do a lot of backend stuff. So you may be responsible for maintaining databases or building data pipelines or other kinds of low level stuff. On the other hand you may be required to apply a lot of greatest hits techniques like a linear regression, or a clustering, or sentiment analysis. Some data scientist positions are very narrowly focused, maybe looking at problems like facial recognition, or translation, and really just tweaking it and iterating it slightly to get the accuracy from 97% to 98%. You really have to look hard to understand what they’re asking for. The other end of that is lots of companies don’t have well-established data scientist hiring pipeline. You may be one of the first data-scientist to arrive in organization and in that case the tendency is that people will ask for all kinds of skills that they don’t really need. For example, not everyone has to understand Apache, which is the baseline open source technology for distributed computing -- breaking a problem into pieces and solving it in parallel; maybe a few years ago you really needed to understand it, but now there are software packages built on top of that really abstract away some of that. You have to read job descriptions really hard, start applying for the jobs and asking what you’re expected to be doing there. One of the big bottlenecks with my work at the UM was data access: data is very political, it may be earned by different business unit or different department. It’s not always nicely prepared in a .csv file ready to be analyzed. You have to develop communication skills, to understand why you should have access to that data, what value you’re expected to bring from it. You also need to be aware of privacy concerns, security and how to think about people personal data. I argue strongly that data scientist should something like a code of ethics ? to really understand what it is they are doing with people data, what kind of biases could be introduced to their analysis, any dangers of re-identification of people even in data that is supposedly anonymous. There is a lot that you can read about it, and if anyone is particularly interested I would refer them to my blog or to message me in person. That is an interest that I developed a bit more for my most recent role, which was working for Facebook in Silicon Valley. The idea of this role is to understand how and when we should use people personal data. We could do all these amazing things like to understand where we should send aid packages after the natural disaster, or understand how we should tune communication strategies if we are trying to encourage people to vaccinate their children, based on what kind of issues are likely to influence them. Are people refusing on the grounds of religion, are they refusing because they don’t trust the government, are they refusing because they don’t believe that it’s effective? This things that we can learn we can then imbibe and inform communication strategy. That brings me to the present day: I’ve recently founded company called Data Apparel. Data Apparel is a line of clothing that uses the power of empathy and data visualisation to better understand the world that we live in. This is inspired by few different pieces from that experience that I’ve described. The first aligns with one of the essential skill of any good data scientist which is communication. The communication can be often burble, but very often visual, big pipeline of analysis can often be condensed into one graph or one figure; it can be shared in the report or on the web-site or in interactive data visualisation. That’s something that any good data scientist should get good at. There is a lot of resources online, lots of great packages like D3 or data visualization experts like Edward Tufte who I want really recommend to anyone interested in this area. I’ve spent a lot of time trying to understand what people can understand, what get them inspired and make them engaged with the finding. The line of clothing that we have takes data visualisations from development data, from what’s happening in the world around us and puts it together in a way that supports a conversation and allows a global citizen to engage with data about the world around us. The other piece that inspired me to start a Data Apparel is the ongoing trend for entrepreneurialism, start-ups and the emphasis, particularly in the Silicon Valley area, that you should start his own company. I really admire that attitude that inspires people to do it, but at the same time, I was very much aware that I don’t have a business background, I don’t understand good business practices and I don’t have any formal training in it. So rather than starting a company that tries to commercialize new cutting-edge scientific methods, I wanted something that have few ??? and bult on core skills that I have; that also will present itself as option to lots of technical graduates: should you join a company or should you start app on your own? That’s just a word of warning: it can be quite a long road to commercializing this work and being able to articulate this value. When you’re in a business world outside of academia you are not always rewarded for doing something that’s interesting or new: people are very pragmatic. Their interests are how can we attract more users, how can we increase sales, how can we make sure that people having recommendation for products or services that they really want and which they really subscribe to. That drive for pragmatic questions makes you nostalgic for your PhD times, when you were able to pursue really interesting questions and think very deeply and broadly. I’ll finish up by saying to anyone who is struggling in the middle of the PhD right now and can’t see light at the end of the tunnel, just be grateful for what you have and where you are now.
That’s my story about finishing my PhD and becoming a data scientist and an entrepreneur. I’ll wrap up a little and say what I think will be very useful to the people who come to the end of their PhDs. Most people will have a tough time in the middle or in the end of their PhDs and, probably, thinking about leaving academia forever. The first point I would make it to reflect and be grateful for the fact that you have this wonderful learning environment, this freedom, you’re able to stretch your mind, because that doesn’t always exist outside of academia unless you end up in an R&D role in industry. Second point is that there is a lot of ?? hype around data science and big data. People might expect that you can do everything, and you might think that a good data scientist has to be able to do everything. But it is not true, a lot of simple methods that you read about are the ones that are used the majority of the time. It is very tempting to memorize huge textbooks of linear algebra and artificial intelligence, but it is not always necessary. Those things are tested at your interviews, but you are really aren’t going to use them on your day-to-day. The third thing is that very often the questions of interests in business at the first blush sound less high priority, than some of the questions you might have come across in your research, pragmatic things about increasing sales or profits. Something you have to come to terms with is that data science is a commercial venture in industry, but at the same time the pay-off that you get is that sense that what you’re doing is going live, it is being scale out to all of the people who use your service. It is the difference that you can make, very quickly; it doesn’t take 18 months to be published in a journal or it’s not some code that is never run ever again, it is live and happening, it’s really nice.
Those just a few final thoughts, I hope they are useful. I hope my story has been useful and it’ll make someone out there think they are not alone. My biography is available and there are links to contact me, so I welcome anyone who has any further questions.
That’s been a wonderful experience. Thank you for having me and letting me tell my story!