Miguel, a pleasure to meet today, what inspired you to co-found Signal AI?

During the years I was doing my PhD, I realised that I liked to do Applied Research, not just pure research. I liked having a problem to solve, like a challenge a user was having. Seeing how research could be applied to a real-life situation was very important to me from the beginning. During the last year of my PhD, I decided that once I finished, I wanted to do something real in the industry, and if I could build something from scratch, even better. I wanted to be a part of a larger idea, not just the product, and be able to apply research using Machine Learning and AI with a clear goal in mind.

Around that time, David, the CEO of Signal, messaged me on Meetup.com and said that he had a well-funded project and that he was looking for a researcher to join him as Co-Founder and that’s basically how the company started. I can expand much more on the history of the company, but what inspired me, was the idea of applying research to real cases, creating an intersection between industry and academia. That has been the main focus of my professional life until now.

What caught your attention about Signal Media in comparison to other options?

Imagine me, a third year PhD student, starting to think about what I wanted to do afterwards. I knew I wanted to create something from scratch, and that I wanted to apply my learnings from the PhD in a practical way. I decided to go meet David and Wes (the third co-founder of Signal), and after talking to them, it seemed like a cool opportunity to be a part of that founding team. The business idea was clear to some extent and I thought we could work well together. From there, we ended up in a garage in North West London working on our goal to make sure executives could read the biggest news stories that might affect their businesses.

We targeted the C-suite initially and, to validate our market, David interviewed lots of C-suite executives in the UK. I think 35 total, and all of them consumed news in a very inefficient way, mostly getting their PA or their assistant reading communications for them, or by reading 5 different magazines at a time. The major decisions in huge companies in this country were being made on that basis, and we thought that if we could help them make better decisions based on external data in a more efficient and effective way using technology there would be a market for that.

How is AI used in the research and data science function?

What we do is try to help companies make better decisions based on external data. In order to do that, we collect text-based media data (including news, broadcast and television transcripts) and we use AI, machine learning and NLP, to get as much knowledge as we can from each piece of information.

For each one of the 4 million documents we process every day, we detect articles about specific companies. For example: is this article about Apple the company? Or apple the fruit? We can identify keywords about our customer’s competition, their own businesses and topics they care about through topic recognition.

We also do quotation detection, so we can detect when people are being quoted, through both direct (using quotations) and indirect quotes. Indirect quotes are very important when you want to understand how and when people are talking about your business or spokespeople and how they’re portraying your messaging and brand.

We also have a research team working on many other fields including areas like clustering or media perception. The latter is a common challenge for many companies and involves using machine learning and AI to identify how the media is describing specific brands or topics and how this representation changes over time. In general, the team looks at our user’s challenges and identify ways to solve them that will be hugely impactful for them.

How do you manage your time between an operational level and Head of Research?

When the company first started, it was growing so quickly, we all had seven hats. I used to do forecasting developing and modelling, basically everything except answering the phones or selling. We all supported a variety of initiatives. As the company grew, we all got more specialised roles.

One of my main responsibilities is to make sure the company is choosing AI principles, ideas and methods across the organisation in the most effective way and to support the alignment between the research, data science and product teams with the company’s overall goals. I do less of the day to day coding now, which is a shame.

As the data science and research teams grow, my role is also to make sure that the team is structured in a way that’s best for everyone, making sure there are clear progression and career growth paths, enabling the team to be the best they can. Now that we have a full-scale suite of executives, I’m less involved in the overall operations of the business, and I’m trying to devote my time to strategic projects that involve AI and technology.

What advice would you have to a Data Scientist looking to achieve the status of Chief Data Scientist / Head of Research level?

Good question! I think the first thing is to figure out what they want to do long term. If you ask anyone in the field, What is AI? What is a Data Scientist? What is a Data Engineer? What is a Data Architect? Everyone has different definitions. I know people who want to be involved in how to apply AI to products, people who want to be managers, and people who want to do day-to-day machine learning themselves. There are many shapes and forms, and they really need to understand their future goals first.

One of the things that we do, is we try to build really clear career paths, some are focused on whether employees want to go higher into a managerial role, and some are focused on helping employees become better at their technical jobs. Neither one of those directions is better than the other, it depends on what people want to do.

In my case at Signal, during the first couple of years, my initial role was mainly coding. I was both a Developer and a Data Scientist. In the third year, I was able to focus more on data science, but I also managed the growing development team in more of a leadership and hiring role. The last year or so was much more focused on strategic initiatives, what’s coming in the next 6 months, not the next week.

What challenges have you faced growing the Data Science function at Signal Media?

Hiring is much more difficult than it looks, as always. I’m sure that’s a common one, as we choose our Data Scientists and Researchers very carefully. Our teams are cross-functional, so we need to make sure we’re hiring researchers that can work closely with product and engineering to make sure that the researcher is doing something that is impacting users – but is also giving them time to do the research their interested in.

The speed and the cycles of development and product versus researchers aren’t always aligned. You can’t do all your research in two weeks, but you can continue doing incremental improvements in the product in two weeks – the ranges are completely different. One of the risks is that you start one research line and maybe two months down the line it’s not a priority in the product anymore. We are always working to align the different aspects of the product roadmap to drive long term clarity.

The other challenge is communication. In our business, you need to be able to communicate with people that are not technical and are not data scientists. You have to be able to explain abstract, complex concepts which is an extremely important skill. Our Data Scientists do this very well, but someone who has just come out of university with a PhD may not have built that skill yet.

How has Signal Media supported the transition from Academia into a commercial environment?

When we hire people who are fresh out of their PhD, we pair them with more senior members of the data science team, and they work with those partners really closely. Apart from that, we work very closely with the product and engineering and development teams so that new employees can learn from them. It creates a synergy, the developers learn more about machine learning and the PhD students/graduates learn more about proper, solid development. We’ve seen that many recent graduates are good coders or developers, but they don’t necessarily know how to put things into production and monitor systems because they’ve had less exposure to that.

Is that something Universities should give more exposure to?

Yes, and no. Coding skills are more important, but some focus on how you can run things in the Cloud, for example, which is really common now, would be helpful.

What work is Signal doing with Universities?

This is something, I have to admit, I’m very proud of. My connection, when I started with Signal, was that I was officially a KTP associate with the University of Essex. The KTP (Knowledge Transfer Partnership) program is a government program to take the base knowledge from academia and apply it to business. They create a bridge from academia to industry. This is something that Signal is hugely supportive of. In fact, at the moment we’re starting a second KTP project with the University of Essex.

Apart from that, we have our Master’s program. Every year we have MSc projects with many universities, including UCL and the University of Essex, where MSc students spend a summer with us, and we give them the complete experience of being a member of our team. They are given a supervisor who is a researcher in the field and they basically continue their research. It’s a win for universities, it’s a win for the company and a win for the student. In fact, many of them continue to collaborate with us over the years or months after they finish the summer program. Many of them have been published after the work experience, we have two publications from students from Essex at the moment and we’re trying to get some from UCL published at the moment.

These initiatives are great for everyone, particularly for hiring and networking. They give us space to work on research lines are likely more lateral to our overall goals and will become more important to us as the business grows over the next year. Often we use these collaborations to focus on projects we don’t have the capacity to explore at any given time. A master’s student is a perfect person to pick these projects up, working here for on average 3 months on a specific problem.

We also have our visiting researchers program which is for people who are already active researchers in their field, usually, second-year PhD students or recently graduated doctors. At those points, often you start being a tiny bit tired from the PhD and having an internship opportunity is very useful. We have had people in the programme from Glasgow, Canada, Madrid, Amsterdam, London, all over.

Do organisations need to do more in this support junior professionals in a commercial environment?

I think for our scale, we’re one of the few organisations that are involved this deeply, in terms of placements for PhDs. I’ve not seen that many, so we’re quite unique in that sense. People tend to have a larger fixed team rather than a smaller team with people coming and going, but we appreciate that, as we always have new points of view. It also allows us to hire people from different backgrounds. We like to think about the diversity of our teams. If all of the researchers on your team are from the same field, they tend to think similarly, whereas if you have people from a variety of backgrounds, you have more ideas.

What did you take from your PhD into a commercial environment?

One of the most important factors was my network. I was lucky enough to go all around the world, and one of the first things that I did when we started the company was contact people in the field and start building my connections. Everyone we hired was a maximum of second connection from someone we knew. Having a good network is key. I’d suggest for PhD students that they go to conferences to network. They’re great for career development and mentorship.

An incredible career so far, and some amazing things to be proud of with Signal, what’s been the best achievement for you and Signal?

The obvious one is that we were three people in a garage in North West London, and when I say garage, I mean garage. In one of our first presentations with a Venture Capital firm, while we were pitching him, his car got an MOT – that’s a real story.

If you think about those three people dreaming about where this company could go, and what we were going to create, and then fast forward five years and we have a company that is doing well, with offices in London and New York with 90 plus people, that’s really impressive. We have a good reputation as one of the applied AI and machine learning companies who are doing real AI and real product development and I’m proud of that as well. We have managed to have a really good product, with strong research and tech, and we are selling it successfully.