For a company that is all about skills, it’s no surprise that TechWolf has adopted a largely skill-based approach to work. When work is done, the person tackling it is flexibly determined based on skills rather than predefined job titles. In the engineering team, this is especially apparent: our job titles feature a low degree of specialisation, even though people in the group often have very niche expertise. Because we are a young company, this comes about organically: people get a comprehensive onboarding, following the rapid development of specific skills where their talent lies (resulting in a typical T-shaped profile for most).

To fill every skill gap in our team, especially under the pressure of rapid growth, we cannot just depend on this general onboarding flow. Even though we can build most of these skills internally, we increasingly need to hire specific skillsets. This brings a challenge: if the scope of the role doesn't follow directly from the job title, that title isn't suitable to advertise the position either.

In the labour market, especially online, job titles are still king: they are the way to summarise a position into a single phrase, helping online job seekers wade through pages of vacancies efficiently. If the title doesn’t fit what they’re looking for, it’s unlikely potential candidates will give the rest of the posting any attention. In some cases, we even see that vacancies barely contain any information outside of the job title. For this reason, TechWolf takes a particular research interest in them. In recent years, we’ve published research on the inherent structure of job titles as a language and how we can model the meaning of titles from their respective skills.

The same research has also resulted in our Title2Profile model, which takes any job title and infers the corresponding skills. It’s robust to languages, misspellings and the sometimes endless creativity people have in describing a position. This machine learning component is applied broadly within our system, as job titles tend to pop up in almost any customer dataset we handle.

However, in our skills-based recruitment, we found it very useful to turn this model around: instead of mapping a job title into skills, we map the skills we’re looking for to one of the hundreds of thousands of job titles out there. The process starts with the relevant team lead describing the tasks and responsibilities we need to cover and specific skills of interest. We infer the corresponding skill profile by passing this on to our very own Skill Engine. Only then, by having the engine select the optimal job title, we name the position.

We’ve found this approach to work very well for our talent inflow: since the name of the position covers the contents now, we attract candidates with much higher precision than we did beforehand, leading to a faster time to hire. Cases like this show that working with skills can help companies close and avoid their skill gaps more efficiently, but also that there is value in connecting your skills-based approach to other models. This way, you enhance processes that fall outside of your skills ecosystem rather than disrupting them.