Ivo Bellin Salarin
July 1, 2025 at 5:00:00 PM
A recent engineering effort highlights how building resilient, API-compatible data connectors and modular dbt transformations can unlock significant value in large-scale ETL pipelines, far beyond simple data movement.
Candidates selection for a position may feel like an overwhelming challenge (Oh! so many candidates applied at the same time!). So you may be tempted by using ChatGPT or other stuff to filter out candidates. (**Beware: this is not legal, at least not in the EU**). I am going to tell you why this is a bad idea!
- ChatGPT is cheesy. Back to the basics, it's a probabilistic text generator, so it will be influenced by the CV text. But their creators superimposed an optimistic take on life.
- ChatGPT is not sure 100%. It can spot a word in a CV (say: terraform) but if the candidate said "I used terraform for testing" openAI will count it as "experience acquired" (even if it's clearly a fake ;))
- ChatGPT doesn't add up like you do. Let's assume that the CV says 24 years of experience (but then you addition all the relevant experience items and get only 12).
- ChatGPT does not transfer that emotion that you experience when reading a CV. It's in your gut, and it's probably the best thing that you can extract from a CV.
- Certain trainings are worth mentioning on a CV, certain aren't. A human with experience knows the difference.
- Every candidate is different. If you were to compare them, you would probably missing elements that candidates did not highlight both. You get an idea about what the candidate is strong at reading the entirety of her/his CV.
What about the law? The EU AI Act classifies this task as high-risk. This implies a certain number of mitigation mechanisms that you should be putting in place to make sure that transparency, explainability and bias mitigation meet standards. Human supervision should be in place. And, you should be paying attention to where the candidate data is being sent (is this a publicly hosted service? Is this a self-hosted AI model?)
The GDPR then obliges you to 1) have a lawful basis (your business interest must outweigh the candidate privacy interest) 2) inform the candidate about how you are going to process their data, and 3) get a mandatory Data Protection Impact Assessment (DPIA).
I hope this helps make the right choice!