Spanish Startup Uses Deep Learning to Match Developers to Jobs


Madrid and Berlin-based Source{d} uses deep learning in a talent acquisition system to identify programmers and other technical talent from open-source projects. Founded in 2014 by Eiso Kant, Jorge Schnura, and Philip von Have, the 2-year-old company says it is already close to being profitable and on track to close the year with something approaching €1 million in revenue. Source{d} claims to have 200 clients ranging from small to medium-sized startups and large corporations. The company recently raised $6 million and plans to expand to London, Paris, Stockholm, Amsterdam, New York, Boston, and San Francisco.

How It Works

Source{d}’s talent acquisition system trawls open-source contributions on GitHub to find out which languages and frameworks the customer uses most. The company then uses this data to identify the best developers to recruit. Currently, the system has analyzed more than 16 million open-source projects with 900 million lines of code contributions from over 6 million developers. Source{d}’s system uses deep learning, so it is not limited to identifying code quality or a developer’s ability; it can also detect other nuances that differentiate one developer from another. The company even claims that it can find people whose coding styles are similar to an existing coding team. Once technical talent is identified, Source{d}’s own developers conduct interviews and technical tests to screen the candidates. So the system is really computer-aided recruiting, as opposed to automated recruiting.

Why There Is a Need

Finding the right technical talent is a challenge. According to a study by CareerBuilder, at least seven of ten information technology (IT)-related job openings went unfilled for longer than a month. The Bureau of Labor Statistics projects that 1.4 million computing positions will be open, with only 400,000 computer science grads (although this figure is a little misleading, as many programmers are self-trained). Technical talent is hard for recruiters to source because it is so specialized that most recruiters do not know what questions to ask to qualify the candidate. Consequently, recruiters can waste a lot of time with the wrong candidates.


A talent acquisition system such as Source{d}’s has many advantages, including:

  • Speeding up the recruitment process
  • Not wasting the candidate’s time
  • Not wasting the recruiter’s time
  • Finding candidates who are a better fit for the company
  • Encouraging people to contribute to open-source projects as job auditions


A lot of coding takes place outside of the open-source community. Even within open-source projects, only a few people do most of the work. According to researchers, such as Tadeusz Chełkowski (2016), “…the participation of contributors is following a steep power law distribution. It is worth noting that open collaboration communities in general follow the ‘1-9-90 rule,’ under which only 1% of community members actively produce content, 9% are generally somewhat active, and the remaining 90% are passive lurkers.”

This means that a lot of talent may go overlooked. It also makes it harder for new, skilled people to be hired for their first job. At the very least, it raises the bar and forces a would-be programmer to learn open-source skills that may not be marketable.

What the Future Holds

Once the technology has proven itself, it can be applied to hiring freelance writers, marketers, and attorneys. For example, a machine vision application could identify fashion models. Or a voice recognition application could choose speakers and teachers.


Companies like Source{d} provide more evidence to support Tractica’s position that the business software sector is poised to leverage artificial intelligence (AI) extensively in the future. Our Artificial Intelligence for Enterprise Applications report forecasts that spending in this sector alone will grow to more than $3.1 billion by 2025, which is a bright future for those entering the market.


Comments are closed.