Note: Please always remember that these tools are experimental and that therefore you have to carefully monitor the result obtained in each case. This is only a proof of concept of the new technologies that are appearing trying to apply them to a real life need.
This summer I had the opportunity to start playing with a new tool from Google, MakerSuite, an interface that allows you to explore the possibilities that the PaLM API offers through a simple interface.
I’m not an expert in artificial intelligence (my thing is the cloud 😜) but I wanted to try something that could be useful to people and it occurred to me to test language models as an assistant in the preparation of an interview process.
This is the main interface if you don’t know it (unfortunately it is not yet accessible in all countries):
The first step will be to use the Text Prompt feature and give it as much information as possible about the job offer in question for which you want to apply and ask it to extract from that job offer a list of 10 questions that could be asked in technical interviews. The more detailed the job offer, the more information our tool will have to know what they could ask us about. I have used this Data Scientist job offer at Seedtag as an example. My prompt will be “Provide 10 technical questions that could be used in a technical interview for this job offer: (full job description)”
And I think the list of questions that it returns to us is already a great starting point:
- What is the difference between supervised and unsupervised learning?
- What are the main types of machine learning algorithms?
- How do you evaluate the performance of a machine learning model?
- What are the challenges of working with big data?
- How do you deal with missing data in your training data?
- How do you handle outliers in your training data?
- How do you prevent overfitting and underfitting in your machine learning models?
- How do you deploy a machine learning model in production?
- How do you monitor a machine learning model in production?
- What are the latest trends in machine learning?
Now that we have some questions to practice, let’s practice our answers! I decide to start with question 4 and I will create a new prompt telling: “Given that a candidate for this job offer: (full job offer description) during the interview answers to the question ‘What are the challenges of working with big data?’ with this answer: (your full answer). How would you rate this candidate?”
Imagine that I decided to provide a short answer like “One of the challenges is that you need a lot of infrastructure to handle it”, let’s check how good I would be for that offer:
We can see that maybe I wouldn’t be considered a good fit because I didn’t provide a detailed answer and it helps me to understand how I could have improved my answer!
And with that we can start automating all this process to try to improve our interview skills, but that would be in a future post if you like it!