Two weeks ago, I was one of the presenters in the launch of the AI Playbook created by David Kelnar and MMC (disclaimer: MMC is one of our investors). The playbook represents a holistic view of all aspects an organisation needs to consider to use and take advantage of AI. It is aimed to the technology-savvy business people in the startup and scaleup community worldwide.
Successful AI adoption in companies will probably be the major distinction between successful businesses and the rest in the coming years. My talk focused on talent and how it holds the key to unlock the power of AI. The talk was received extremely well and I thought it was worth sharing my learnings (and my mistakes) over the past 6 years as the Chief Data Scientist (and technical co-founder) of Signal AI. In this time, Signal has grown from three people in a garage with an idea to a company with hundreds of clients, offices in three countries, employing 150 people, including a full time research group of nine and numerous academic collaborators visiting us at any given time.
In this series blogpost I will focus on how to source the right people, how to assess their fit for your company and the role, how to structure the data scientists in your organisation and finally, how to retain and develop this talent. I will address each of these three aspects in separate posts but, for those of you who cannot wait, these are my main take-aways for the whole series:
- Make yourself known in the community.
- Hire for more than just technical expertise.
- Make sure you have a clear problem that AI can support.
- Understand and take care of your Data Scientists.
Most importantly, remember that a company is nothing without its people.
The first challenge related to talent is to know where and how to find talented individuals and in the case of Data Science, the competition is even fiercer than in other areas.
"Data Scientist is the sexiest job of the 21st century" - Harvard Business Review (2012)
One of the most straightforward solutions to find talent is, like in other fields, through recruiters. Like everywhere, some recruiters are better than others and you will find a wide range of fees (usually between 10-20%). My suggestion is to always go with a recruiter who specialises on AI / ML / DS as the candidates are difficult to navigate and assess and for recruiters who are not fully specialised in the space. Ideally you will have an experienced in-house recruiter from the field who is carefully guided by your data scientists in terms of the profiles you are looking for. However, this model is only suited if you are planning to grow your Data Science function significantly (i.e., at least one new hire every two or three months). Otherwise, the cost and complexity will not be reasonable.
A complementary approach is to attend meetups. I have made many great connections over the years in different meetups in London (e.g, PyData) and abroad and have always learnt something new. Just as an example of how impactful meetups can be, the three co-founders of Signal AI met when David (our CEO and co-founder) “spammed” both of us using two meetup groups and we agreed to meet to discuss the initial idea that will become Signal AI. Without meetup.com, the Signal we are today would not exist. The downside is that they can be a hit-and-miss because it is very difficult to predict the impact and value of each one, especially for meetups outside your area of expertise. Another challenge is that more meetups tend to be located in popular urban areas (e.g., London) while they are not as common in other geographies (e.g., my hometown of Oviedo in Spain).
A similar approach to meetups is to attend conferences in order to increase your brand visibility and to network, which eventually will drive hiring. My advice is to attend both academic and industry conferences, even if you are not hiring now. Company branding and reputation takes years to build but it is the most powerful way to ensure great candidates will come to you and recognise your name. In the case of Signal, we have been attending and speaking at many applied AI and ML events in the UK over the past years. From an academic point of view, we have been very active in some of the communities (e.g., Information Retrieval and Text Analysis) through publishing papers and by co-organising multiple events. In 2016, for example, we organised the First International Workshop on Recent Trends on News Information Retrieval (NewsIR) as part of the ECIR conference. The reception and the event was fantastic wand we ran a second edition of the workshop in 2018 as part of ECIR’18. The third edition is about to happen next week, this time in SIGIR, the top international conference on Information Retrieval. These events allow us to influence and focus multiple lines of work that the community has been working on around topics of interest for Signal AI, while being able to share our industrial and applied knowledge to the community. As a result, our recognition and network has increased remarkably. In addition, this event allowed us to work with other co-supervisors from a range of academic institutions and other technical companies such as the University of Sheffield, Columbia University or Bloomberg, just to name a few. All these factors have been critical to shaping the perception of Signal as one of the best places to apply proper research and having impact while solving real problems.
One of the main approaches for Signal to attract talent has been through our university collaborations. Some of the best research is still done at universities and I strongly believe that the more academia and industry collaborate, the more innovative we will be as a community. Our research quality, publication record, reputation and hiring have benefited massively from external researchers visiting Signal and MSc students doing their thesis with us. In fact, two of the MSc students we recently co-supervised (from UCL and Essex University) became full members of the team. Universities are open to negotiate the details of collaborations with industry and you should be very clear about IP and commercialisation rights related to any project from day one to minimise any misunderstandings. A major point on talent is to focus not only on the university name but to focus on the lecturers and researchers and their academic contribution in your area of expertise. We have seen first-hand how reputation and referencing in a small community is a powerful factor, even when dealing with a worldwide distributed force. Many of the people applied to Signal because of our reputation in the community and/or the word-of-mouth from Signal alumni (including the academic collaborators). This leads to one of the most important realisations about hiring: the AI community is relatively small and your reputation will precede you, so it is absolutely critical to be always fair and transparent. One of the goals for us now is to expand our reputation worldwide and to be much more influential in the more general areas of Machine Learning and Artificial Intelligence communities.
Don’t wait until you need to hire at scale to build your own company brand. Create a name in your community from day one, then plan how to scale that reputation.
I hope you have found this piece interesting, and if so, I will be releasing the second iteration of the series next week.