Solid interview, thorough and fair, interview was conversational and centered around broad concepts in AI and ML. Interviewers were open to bounce ideas back and forth, and there was interest in going deep into one particular topic or method.
Interview questions [1]
Question 1
How would you design a RAG system? What will impact performance the most?
Badly structured, pretty chaotic startup without product market fit and weak open source project. Cofounders did not seem very technical just doing their pitch without knowledge. A deep sense of cluelessness from engineers
Interview questions [1]
Question 1
Basic Machine Learning questions, average difficulty
I applied online. The process took 4 weeks. I interviewed at MindsDB in Jun 2020
Interview
Three rounds:
1. Machine learning questions, easy to medium difficulty. Mostly conceptual, and some applied stuff.
2. Paid take-home exercise from one of their ML-heavy open-source GitHub repos. My specific problem was quite hard, and took me over a week of part-time work. Compensation for this was adequate, and in the end (as I got the offer) my solution was eventually merged as a product feature, which I found cool.
3. If your solution is good, a third round would have you closely discuss your approach with an ML engineer on their end. Trade-offs, alternatives, behavior, results, etc.
Interview questions [3]
Question 1
Implement an autoencoder RNN architecture in PyTorch, able to reconstruct input time series and also forecast (t+1, t+n) future values.