Disclaimer: All opinions expressed are Alex's and not those of his employer.
Hi Data Friends!
I have just had the pleasure of interviewing the awesome Alex Antic. Alex is a Data Scientist and Mathematician who has had an absolutely stellar career.
Alex was kind enough not just to tell us a bit about himself, but also to give us some real insight into the state of Data Science and Analytics in business and government.
Enjoy his great tips, experience and advice!
1. Your credentials are incredibly impressive. Can you tell us about how you got into data science and your journey from Wollongong to Canberra?
I guess it's fair to say that my journey began, many years ago now, with my studies in Pure and Applied Mathematics, Statistics and Computer Science at Uni. The combination of the two degrees (Mathematics and Computer Science) allowed me to combine my passion and aptitude for math's, with my interests in computer science, as I saw coding as a practical and powerful way to help solve complex math's problems.
After a brief stint as an academic after completing my PhD in Applied Mathematics, in which I enjoyed teaching and conducting further research, I decided it was time to enter the 'real world' and to put all I'd learned to good use (ie I was fed up with being poor and wanted to finally make some money!).
Along came an opportunity to move to Sydney and take up a role as a Quantitative Analyst (quant) in a Fund of Hedge Funds for the largest fund manager in the country (back when Hedge Funds were sexy and all the rage). It was an interesting role, where I got to apply my skills to the field of portfolio optimisation, looking at the best mix of investments in order to optimise the companies' profits. I was lucky to meet some smart, interesting, eccentric and extremely wealthy people during that time, somewhat of an eye-opening experience, and even met a few famous ones too, including Professor Ross Garnaut, previous Senior Economic Adviser to Bob Hawke, Dr Nassim Taleb, the author of the well-known book The Black Swan (amongst others), and Dr Myron Scholes, co-originator of the famous Black-Scholes options pricing model, and a hero to many quants!
After about a year, I felt there was more challenging work that I could be doing to make the most of my skills and experience, so when I was approached to join one of the most famous investment banks to work as a Senior Front Office Quantitative Analyst, I jumped at it! There I spent almost six years, honing my derivative pricing, coding, mathematical and statistical modelling skills, to price complex financial products. I was lucky to work with some of the best quants in the country, for a fantastic, supportive and dynamic company, and successfully completed some challenging and exciting projects.
Ultimately though, I felt that I’d learnt all I wanted to learn as a quant, and felt somewhat worn out from all the long hours, high stress, demanding work and effectively no holidays. Coupled with the fact that I was yearning for a more senior role, that would allow me to take on further management and mentoring opportunities, I made the difficult decision to leave, and to take up a role as the Manager of a Marketing Analytics team for one of the world's largest insurance companies. It ended up being yet another great opportunity, as not only was I still hands-on with the tools, but I was managing a team of skilled analysts and programmers, and helping develop their skills, while doing some leading edge analytics work. This was actually the first time I dabbled with Machine Learning. A part of my role was also as the Lead Data Scientist of the Australian arm of the organisation, which meant I assisted various analytics teams throughout the company on niche projects, one of which was to assist the actuaries with their transition to Machine Learning valuation algorithms.
However, after having spent over 8 years living in Sydney and working in the financial services sector, I was feeling somewhat disillusioned. Not only did I feel that a big bustling city was not the place for me any longer, I was also struggling with the morals and ethics of using all the skills and knowledge I'd acquired to simply help make rich companies even richer!
So when an opportunity arose to move to Canberra and take on a senior leadership role, and become a fully-fledge Data Scientist, with opportunities to manage the team and help set the strategic direction of the branch, for one of our biggest Federal Government Departments, I couldn't resist! I finally had the opportunity to achieve a more favourable work-life balance, and to use my powers for the greater good, by helping make our country a better and safer place (as corny as that may sound).
To succeed in your career, you need to be flexible and agile like water. Go with the flow, don't let obstacles in your way stop your forward progress, and expect and be prepared for turbulence.
2. Canberra is a great place. Can you tell us about living there and what you get up to away from the desk?
It sure is Nic, and I feel privileged to have the opportunity to live and work here.
First of all, it gives me the opportunity to work in the Government sector, and to play an influential role in helping develop the analytics and Data Science capabilities of our Government agencies. It's something that I'm passionate about, find rewarding and take pride in.
Secondly, I feel extremely fortunate to be able to live on a rural property (something that's been a dream of mine since I was young) and yet be only half an hour away from work! What's more, I live near a national park and the Australian Alps, that are snow-capped in winter, with rivers nearby that provide great swimming spots during our hot summers.
Whereas I used to be woken by police sirens, neighbourhood disputes and traffic noise when living in Sydney, I'm now greeted by the sights and sounds of sheep, cattle, birds, kookaburras, kangaroos, wallabies, echidnas, deer, and friendly possums and wombats that greet us at the back door most evenings to say hi and request some food!
All this allows me the exciting and relaxing opportunity to spend as much time outdoors as possible. This includes road cycling, mountain biking, trail running, horse riding, hiking, camping and photography. There's always new places to explore and adventures to go on!
How cool is Canberra?
3. You have led data science teams and projects, you have a huge amount of experience across different industries. What advice can you offer to data science managers about nurturing new data scientists?
Great question Nic, and one I get asked often, especially when assisting Government, industry and start-ups with setting up Data Science capabilities, and then knowing how to effectively manage them. It's an important question to address as I've seen too often many failed attempts, which have resulted in the loss of great people.
I recently discussed some of my views on this very topic during a podcast interview (amongst other topics). The link also includes an infographic highlighting some key points that readers may find useful.
In summary, I feel the following points are worth noting:
Data Scientists require different needs than more 'generalist' employees. For instance, from a technical perspective, they need suitable hardware, access to specialist tools and software (typically open source) and access to data and key stakeholders. In order for them to be able to truly understand the business, before they can come up with innovative technical solutions, they need to first learn about the business directly from the SME's, who can also assist them understand the intricacies and complexities of the data, and the business rules that govern them.
Hierarchy tends to be much less important in Data Science teams, and can in fact make things much more difficult. Flatter structures tend to work better as it gives them an ability to engage with peers and stakeholders on a more equal footing.
Keep them engaged and challenged! It's no use hiring smart people to only have them work on mundane work that less qualified, less knowledge-thirsty and cheaper staff can do. They enjoy a challenge, and to be rewarded with praise for their success, and transparency with how their efforts are being used. Adequate salary and career progression are obviously important too. They also value transparency of who the main stakeholders are, where they can access data and information they need, and assistance in gaining access to such information.
When it comes to management style, definitely don't micromanage them! They tend to prefer autonomy and to be trusted to ask for assistance when needed.
They greatly enjoy the ability to have research opportunities in their roles, scope to try new ideas and test assumptions, and people to bounce ideas off. Access to technical and thought leaders, opportunities to attend conferences, Meetups, and technical forums, and to present at these, are valuable to both them and the organisation.
Metrics are important too, so both employer and employee can measure the success of generating meaningful and actionable insights. These include measuring the value of the projects (such as money saved, generated etc) and the number of projects completed.
For those Data Scientist that are starting out, mentorship is important. Try pair them with more senior technical staff and encourage them to work on more advanced projects alongside these staff to help them learn and keep them challenged and engaged.
Given that analytics is an iterative, non-linear, process of knowledge discovery, it needs to be managed as such. Managers need to give their team opportunities to operate in such a fashion and not to be scared of making mistakes, as often, valuable knowledge is gained in such ways. However, you want any mistakes to be made as fast and early as possible, and as cheap as possible, both in time and money.
An important aspect for aspiring and junior Data Scientists is for them to have freedom to ask questions and confidence they'll receive the answers they seek. There needs to be a process that allows freedom for asking questions, and seeking appropriate answers, with support at all levels.
It's also important to give them opportunities for further study, and for them to connect with others at technical and business gatherings.
In regards to managing Data Scientists in general, make sure that they're not isolated! They need to feel engaged and be given an opportunity to gather and share knowledge. The best Data Scientists tend to work in teams with other Data Scientists, and not in isolation from one another or from the business, unless they're in a specific research role.
In general, I believe it's best to have a centralised analytics team within an organisation, where possible. Alternatively, I think it's important for managers to work together to ensure that analysts from different teams collaborate to share ideas, skills and experiences. This is for the greater good of both individual analysts and for the organisation, as there's no point in infighting of for any team to re-invent the wheel.
I also believe that analytics teams often are most productive when they are managed by technical managers (who also have great business and communication skills). Such managers tend to understand the technical aspects of the roles they’re managing, the motivations of their staff, and their specific needs and requirements.
4. While we are talking about new data scientists: we have companies screaming out for data scientists, and yet we see a lot of new data scientists struggling to get their first start, what the heck is going on?
I agree that there exists a perception that this is the case, and it may very well be the case in some industries and some countries, but within the Federal Government sector, I don't think it's actually true. The reality is that there are often many more candidates applying for a role than there are roles to fill. For instance, a peer of mine who recently recruited for a single Senior Data Scientist in their team for a Government agency, had almost 60 applicants, but only a small handful were sufficiently qualified enough to warrant interviewing.
Given the large growth in the popularity of Data Science, and the number of people either switching from related fields or entering without a formal quantitative background, there is a lack of qualified Data Scientists at the senior level for employers to choose from. And for aspiring and junior Data Scientists, they may find that it's a competitive field as both industry and government are still relatively immature in their analytics capabilities. I think it will take a while yet before we see a significant growth in the number of true Data Science jobs on offer.
Some of the confusion also arises from companies not knowing what they want or need. Too many are jumping on the band wagon without a clear directive, lack of support or specific strategic goals to ensure the success of a Data Science practice. So they try to hire a Data Scientist or two, but ultimately struggle to keep them through lack of challenging work, sufficient support, or clear and measurable value metrics.
It's imperative for Data Scientists to watch out for some of these warning signs before taking on a new role. Here's a couple of articles I've written to help aspiring and junior Data Scientist navigate this tricky area:
In this article, I discuss the 10 hidden challenges of working as a Data Scientist, and how to overcome them to advance your career; and
In this article, I discuss the 7 questions you need to ask before taking on a Data Science role.
Another issue, from the employer’s perspective, is not having a clear idea what Data Science really is, and what it can offer. They ultimately need commitment, from senior management, to strive for a data driven decision making agenda, and to have clear goals of how this ties in with their business strategic goals. Once they figure this out, they then often need assistance in building up the capability, knowing where to place the Data Science team, how to get engagement from the business and IT, and finally, what skills to look for when hiring Data Scientists.
For Data Scientists struggling to get their foot in the door, I can't stress enough the importance of both mentoring and networking. Apart from that, it's worth keeping an eye out for Government agencies that often have entry level roles, or internships, as do some private organisations, that are willing to take on junior and aspiring Data Scientists and help train them up. It's also worth trying to find internships and scholarships during University summer/winter breaks to get some practical experience.
Another helpful tip is to build up a profile and some practical experience via GitHub, personal blogs, LinkedIn, and having specific projects that you've worked on to discuss in interviews, which clearly state the value you developed and the technical skills you demonstrated.
5. You are keen on networking, tell us about your involvement in the data science community including Canberra R Users Group, Data Science Canberra and Analyst First – Canberra and the benefits of networking with the data science community through meetups, LinkedIn and conferences.
My passion is to share my skills and experience, to help mentor and educate Data Scientists, and to help connect them to like-minded people.
However, I didn't always find the networking aspect easy. Being a Data Scientist and a bit of a nerd, means that I tend to be an introvert and somewhat shy at times, so used to avoid networking. I've had to learn to work through this as I felt it was holding back my career in the early years. I still struggle at times these days, but feel a lot more comfortable after realising that most people genuinely want to learn and share from one another, because after all, as humans we're social creatures.
We're so lucky these days with sites such as LinkedIn, which have a huge pool of likeminded and talented people, such as yourself, who are willing to share their skills and experience and offer valuable experience to others. I also find it to be an invaluable resource in landing new roles, as I'm often contacted by headhunters both locally and abroad for new and exciting opportunities, so I don't tend to have to formally 'apply' for roles anymore.
I often encourage others to either attend Meetups or start their own. They're a great way to speak to like-minded people, learn about new developments in the field, share your own knowledge, work on problems together, and of course network in the hope of landing a new role, or trying to find someone to fill a role in your own team.
The Meetups I run are a great opportunity for me to catch up with peers and former colleagues, and to hear about some of the great work they're doing. I also get a chance to discuss in person questions that aspiring and junior Data Scientist have for me, and I get to connect people together in the hope of helping fill vacant roles. I also enjoy promoting the exciting work being done in the Government sector, that otherwise may go unnoticed. For instance, I'm planning on running a 'Women in Data Science' series soon that highlights the exceptional work being done by some of the lead female Data Scientists in the Federal government space.
As mentioned above, the importance of networking for Data Scientists can often be underestimated. As your career grows, it's often the most effective way to either find a new role or to fill a role in your own team, and it's a great way to connect with people that share similar values, interests and career aspirations.
6. How did you get into consulting/contracting work? What advice would you have for data scientists making the move to starting their own business?
It was something that I'd been curious about for a while, so when I was lucky enough to be asked to build and run the Data Science & Analytics practice in Canberra, for a large consultancy, I thought it would be a great challenge and learning opportunity for me. I was at the point in my career where I not only wanted to still be hands-on with the tools, but to also help guide and assist organisations, and Government agencies, with large transformational projects and niche complex analytics problems.
I wanted to play a greater role in assisting them to meet their strategic aims via the power of Data Science, and to realise all the success that can be delivered via data driven decision making. I felt like I'd learned so much throughout my career, at various levels and in different industries, that I wanted to share that knowledge and help others do it the right way, in in a sustainable fashion.
Simply put, I wanted an opportunity to work directly with clients, at senior levels, to play a greater influential role in guiding them along their analytics journey.
A key element of my role these days is to assist government departments and agencies, academia, industry and startups with developing, from the ground up, capabilities to enable data driven decision making, and to use my expertise to assist with solving complex and challenging technical problems.
In regards to others wanting to move into the area of consulting, contracting or starting their own business, the main advice I can offer is as follows:
You need a strong network. Not only enough people that you can turn to, but a group of people than can vouch for you, assist you, help you grow your network further and offer valuable business opportunities to get you started.
Communication is key! It's vital to be able to communicate technical concepts to a business audience, and to adapt and understand business jargon, so you can translate their needs to technical solutions.
Confidence. Don't be scared of failure and believe in yourself. As someone once said, FAIL = First Attempt In Learning. So try to fail fast and cheap, and keep learning from any mistakes on the way.
You need a proven track record, a way to convince potential employers of your expertise in an area. You need to be able to demonstrate results you've delivered, be able to communicate this effectively, and prove that they can depend on you to make a positive impact on their business. This can include posting code etc on GitHub, your personal blog, LinkedIn and having influential people (such as former managers and mentors) vouch for you. For such reasons, you need to spend time in building up your career and reputation, especially in the Government sector.
It's imperative to be known as an expert in your field, someone who can hit ground running. Typically, you're required to acquire a lot of in-depth knowledge rapidly, synthesize this into an actionable technical solution, then effectively manage the successful delivery of the project, sometimes, completely on your own. When you're a permanent employee, they're prepared to invest time and money into developing your skills, whereas when you're a consultant or contractor, you're expected to effectively walk in and solve all their problems!
Also, be prepared to work on your own at times and in isolation. This can be challenging for some as often as Data Scientists, we work best in teams with other Data Scientists, where we can bounce ideas and collaborate. However, don't let this stop you reaching out for help and keeping the business informed with any roadblocks or unrealistic milestones. Ultimately, they want and need you to succeed (after all, you're costing them a lot of money, and they have a specific problem that needs solving) and will try their best to help you.
7. What was the idea behind the Impartially Derivative blog? Great name by the way!
Thanks, I couldn't resist such a nerdy math's pun :-)
The impetus was three-fold:
It's a great avenue for me to share many of my thoughts, ideas and technical musings in a manner that allows me to engage with others, and hopefully spur discussion and thought.
I'm often asked for career guidance, and specific technical assistance with concepts that underpin much of Machine Learning, Deep Learning and mathematical and statistical modelling, so I thought the blog would give me an opportunity to help explain some of these concepts to a broader audience. Given there are many other blogs that do a great job in explaining similar technical concepts, I wanted to focus on explaining the intuitive components of many of these key themes, especially from a mathematical viewpoint, something that I feel is often missing. I'm passionate about trying to help people understand and appreciate the science and beauty behind mathematics, and to help demystify it!
I'm lucky to have worked with some great people, many of whom have become successful in their chosen fields, and to be connected to a number of thought leaders and influencers that I respect. By interviewing some of these people, and asking questions that I feel would add value to the broader reader, I can not only provide some key insights as to who they are, how they became successful, what a typical work day looks like for them, and their secrets to success, but also promote some of the great work they do. Many interviews exist with the most well-known and successful Data Scientists, so I wanted to instead focus on the fantastic work being done by those equally great but less well known, such as yourself, and the many great people both you and I are connected to :-)
As a Data Scientist looking to grow your career, don't just focus on the destination, enjoy the ride too!
8. I was trying to get access to a data source when working at a credit bureau, one of my mates famously told me “Nic, you can’t spell bureaucracy without spelling 'bureau'”. How is the red tape working in government, and is it getting better?
Great question Nic! Many people outside the Government sector often ask me similar questions.
To begin with, would you believe that some large private organisations actually have more red tape than some of the smaller, more agile Government agencies?
It ultimately comes down to leadership, flexibility and agility!
Some agencies, especially the larger ones, and those specifically focused on national security, obviously have more of what we call 'red tape', which is understandable, given legislative and legacy requirements and restrictions. Most are aware of this and are (slowly) doing their best to overcome these challenges.
However, there are some common challenges and intricacies within the Government sector:
In some departments and agencies, you require a security clearance, which can take up to 12 months to acquire.
Data classified at the Protected level (or higher), is stored on air-gapped platforms at a minimum, with no internet connectivity. Data Scientists who are transitioning from academia, for example, will need to adjust to not being able to quickly and easily experiment with new tools and download packages.
Data sharing between agencies can be a challenge. There is some great work being done to improve this, but often the problem is aging legislative frameworks rather than technical limitations. A recent successful data matching project between the Department of Human Services and AUSTRAC shows what can be achieved in this space. This is just one example of some of the incredible work being done in the Government sector, especially by the talented, successful and respected Data Science team at AUSTRAC.
Procurement processes and funding limitations can result in long lead times for hardware upgrades.
Some agencies have limited understanding of, and maturity with, Data Science. Such agencies may have non-technical generalists leading teams of Data Scientists, or even have a single Data Scientist tasked with setting up a capability. Both of these situations present challenges to the Data Science practitioner, even when they're a specialist consultant/contractor.
Funding limitations, long procurement processes, lengthy security clearance schedules, and various myths, can make it difficult to attract, hire and retain talented staff.
The prevailing political environment has a greater impact on the Government sector than other industries. There can be frequent and significant Machinery of Government (MoG) changes, such as agencies merging, teams being disbanded and staff numbers dropping, that can be disruptive and destabilising, which can sometimes result in relative mature Data Science teams and capabilities being disbanded almost overnight.
The level of Data Science and analytics maturity greatly varies between Government agencies. Most still have a long way to go, at an enterprise level at least, but almost all have pockets of talented staff doing some exciting and valuable work. I believe that such issues are common in industry too, as Data Science as a whole is still in its infancy.
Overall, I find working within the Australian Public Service (APS) to be challenging and highly rewarding.
9. There is a debate with the purists who believe every model should be cut from first principles while avoiding all modules and packages at all costs. Others believe the practicalities of business require us to use open source tools and libraries for consistency and speed of implementation. As a guy who has experience working with implementing algorithms in C through to managing teams and projects what are your thoughts?
Very interesting question! Yes, I started my career where we had to implement all fundamental models and techniques from scratch in C, C++ and Python, something that is basically unheard of these days. Whereas now, I can use libraries and functions where I don’t need to worry about the inner workings, garbage collection or in depth debugging.
As a former academic, and having worked with academics who've also made the transition into either the private or public sector, I always stress to my staff the importance of pragmatism, and the need to focus on achieving results, and not risk wasting time to reinvent the wheel.
We have an abundance of well-established packages, libraries and tools, that are peer-reviewed and constantly being improved and upgraded (the power of open source), so why not make the most of this!
The real key is to understand the fundamentals, at least from an intuitive perspective, as that's what really helps make someone a great Data Scientist. The tools, models etc aren't as important as the reasons for solving the problems and how you go about it. Asking WHY is more important than focusing on HOW!
In practice, we need to be pragmatic and often work on multiple projects at the same time, various administration tasks, and research projects, all whilst managing stakeholder expectations, tight budgets and limited timeframes. I tend to advise people to work in an iterative manner, in that they develop the bare minimum possible to begin with, then work closely with the stakeholders to refine and adapt until the desired result is reached.
Remember, as a Data Scientist you're there to solve problems, and specifically, to turn data into insights to make decision making better, so focus on the business problem first and foremost, then figure out the easiest and fastest way to solve it.
It's often easy and extremely helpful these days when teams have access to peer-reviewed and peer-developed code stored in repositories such as GitLab, which can make it so much faster and easier to create reproducible results, especially for new team members learning in which disparate systems the data resides in, and what complex business rules govern joining it!
10. There is a buzz around deep learning now. Do you see widespread adoption of deep learning models in business and govt as opposed to more standard regression type models happening soon?
It's a very topical point, and I recently wrote an article that touches on this. I think there is great merit in Deep Learning, but it ultimately has a long way to go to before we reach true "artificial intelligence". For that we need models that can incorporate abstraction and reasoning. They excel at certain tasks but are definitely not the Machine learning panacea that some claim them to be, as I've previously discussed in this article.
I think, however, that Deep Learning has already become more mainstream, and will continue to do so. Just look at how much awareness there is in the general media, and the number of technologies that are trying to embed it into their products. I know of many Government agencies, organisations and start-ups that are investing in it, and doing extensive research to incorporate it into their systems. However, as the hype cycle wares off, more effort and attention will be placed on its limitations (as is the case with all exciting new developments) and the next new thing will evolve, including improvements to Deep Learning algorithms, which will be very exciting to see.
The fundamental point for Data Scientists is to focus on the problem, and not the solution. Once they understand the business problem, and the data they are working with, should they choose the most appropriate solution, starting with the simplest, and only then increasing complexity if need be. This will aid understandability, validity and ultimately the capability of 'selling' the solution to senior executives, stakeholders and clients, which can be difficult when dealing with a black box solution.
Before any such advanced models become more widespread, analytics in general will need to become more mainstream, and by that I mean, using data much more broadly to assist in decision making. For that to happen, the C-suite and broader organization will need to increase their data literacy. More education is required to help manage expectations, and to help senior executives ask the right questions and know how to interpret results.
The ultimate aim for us is to enable better decision making!
A Data Scientist deep in thought about Deep Learning