What is an NLP internship at TechWolf like?

A TechWolf Tale by Vic Degraeve

During the summer of 2020, I interned at TechWolf as an NLP Engineer. I was assigned my very own end-to-end NLP project, with a real-world use case. The aim of this project was to discover new skills from an extensive collection of unstructured vacancies. Using this discovery system, the Skill Engine™’s understanding of skills can then be updated regularly.

Working on an end-to-end project, I was in charge of everything from the dataset, to the evaluation strategy, to the comparison of different models. As most of my encounters with data science until this internship had been at university, I only had experience with the latter: building models. While this is undoubtedly important, I quickly learned that a clear evaluation strategy and definition of success are indispensable to any real-world data science project.

Having defined the problem, I was ready to embark on my journey through NLP. It started with a thorough survey of research on related issues. Apart from reading, this often involved implementing models from (vague) descriptions mentioned in research papers. Implementing these models turned out to be more challenging than it seemed at first, considering that any implementation had to scale to a large number of vacancies. Converting papers into scalable code is certainly an excellent skill to have, and I’m glad I had the opportunity to practice it!

After trying out existing approaches from literature, I tried to incorporate what I had learned into custom models tailored to skill discovery. These ranged from leveraging simple patterns manually observed in the dataset to altering tried and true NER models to supercharging old statistical keyphrase extraction methods with fancy new deep learning concepts. In short: I was free to try out whatever I came up with. Fun stuff!


I experienced this freedom throughout the internship. While I had great guidance in defining the project’s goals, evaluation strategy and code framework, with feedback every step of the way, I was still free to discover important insights on my own. I feel that this contributed greatly to my learning process. By experiencing the challenges of building a machine learning project from scratch myself, while getting helpful feedback and tips where needed, I learned so much that I couldn’t have, just from reading about it.

TechWolf was really welcoming to me. I was treated as a co-worker, not as an intern. I got to join in on after-work drinks, go on a fun team-building adventure and experience an exciting move to a bigger office. I felt part of the team. It was obvious to me that the whole TechWolf team is working on something they genuinely believe in, and this really motivated me to go the extra mile.

All in all, if you want to learn something about real-world machine learning while experiencing the thrill of a growing startup full of talented people, you might want to get in touch with TechWolf. I think you’ll love it!

Want to become part of the pack?

Contact us because we are incredibly curious to meet you!