Our current process looks like this:
- Initial Chat (15min): intro, discuss compensation, ESOP, CV...
- Coding Task & Design Brainstorm (1h): A minimal coding task and system design case study. Focus on analytical thinking and approach to problem-solving.
- Take-Home Task (2h): A practical ML/backend/frontend problem
- Final Meeting: Review of tech stack, potential projects for the new hire, and final compensation alignment.
- Decision: Offer or not
I'm interested in your experiences and suggestions for refining this process.
What worked well for your team? What did you like or wish for when interviewing at other startups? Any pitfalls to avoid?
What are you trying to assess with each session? Are you cargo culting or do you really know what you're evaluating?
Is there a part in the process that focuses on selling the role to the candidate?
Not sure you need the take home task, just roll that into the 1hr coding task. With ChatGPT these days, pairing is the best for getting experience with a candidate’s innate abilities. Not everyone does as well pairing as they do solo though, so try to take that into account when evaluating them since they probably won’t be pairing on a normal basis. Pair on the same problem for every candidate and have a specific list of evaluation criteria.
Instead of the 2hr coding take home, get a really broad list of coding related questions together and spend an hour discussing those with the candidate. The idea is to evaluate the breadth and depth of their experience.