A. For interviewers

Candidates are frustrated with the hiring process -- it's beyond their control. But hiring managers also seem frustrated with their own processes because they can’t or don’t know how to change them. Every hiring manager I’ve talked to expressed interest in reading this book to learn from other companies. I don't hope to change a system with one book. I just hope it provides a scaffold on which we can start the conversation and establish good interviewing practices.

Hiring for ML roles is competitive. Strong candidates always have options, but strong candidates in ML have a lot more options. A hiring process with the candidates’ experience in mind will give candidates a better impression of your company, making them more likely to accept your offer. A better process can help you attract more suitable candidates, make better hires, which saves you time down the line. Even if it’s a no-hire, you still want to win their allegiance. You might want to reinterview them in the future, they might recommend you to their network, and they can become advocates for your products.

ML interviews are still a black box. For a field so obsessed with qualitative methodologies, it’s surprising that there has been so little qualitative research done on the hiring process for ML roles. Few companies can state with any certainty the correlation between their interview techniques and the quality of hires. Some of the challenges include:

1. People who interview aren't the same people who evaluate performance on the job.
2. No established metrics for interview outcomes and performance outcomes, or the correlation between them⁵⁷.
3. No control group. Companies don’t hire people who fail their interviews to compare their job performances to the performances of those who pass.

A low hanging fruit is testing your pipeline on existing people within your organization. See how they like the process and whether their performances on the interviews correlate with their performances on the jobs. Another is actively soliciting feedback from candidates. Instead of sending out survey links that nobody clicks on, after each onsite, have someone not involved in the hiring process spend 15 minutes talking to the candidate. Make sure the candidate knows that their feedback won’t affect the hiring decision in any way.

If you don’t have a training program for interviewers already, please set up one. Interviewers are both gatekeepers and faces of your organization. Letting new hires shadow for a few interviews isn’t enough. New hires have no idea what they are doing, and companies have no way of standardizing feedback across teams, interviewers, and candidates.