I applied through a recruiter. The process took 1 week. I interviewed at 1stDibs.com (New York, NY) in Nov 2023
Interview
Process took about a week. You're being interviewed by people with non ML engineering experience for an ML role... Two of the three interviewers were seemed nice, but questions were incredibly vague. Its not their fault though, how could you dig into specific ML related questions with no ML background? Also one guy yawned while I was answering a question and probably forgot he was even interviewing me. If you want to do well on this interview just memorize a model for a recommender system and what input you'd expect for it. Last thing I'd warn you of is take a look at the companies financials (they are public). They are losing ~20 million a year. Their ship is definitely sinking.
Interview questions [1]
Question 1
- Pytorch code explanation
- ML end to end process
- rec system design
I applied through a recruiter. The process took 1 week. I interviewed at 1stDibs.com (New York, NY) in Nov 2023
Interview
They are starting from scratch on ML. No infra to support/deploy large-scale ML deployment, No tech solution for visitor ID stitching. No CI/CD setup. But it seems they are very convinced that ML/AI gonna save their sinking business ship. The interview experience was pretty dull, Each threw one random question and then got really silent. I stopped many times to ask clarifying questions or do I need to go into details. Always ONE word response. one of the interviews seems super distracted as if he got somewhere to go or better things to do. I would avoid this company, it seems the interviewers are looking way out rather than staying put. BTW, all 5 rounds are conducted by white male interviewers...
Interview questions [1]
Question 1
Why 1stDibds
item-item based rec
personalization
explain below pytorch script
I applied through a staffing agency. The process took 2 weeks. I interviewed at 1stDibs.com in Oct 2023
Interview
I was interviewed by 5 white males, none of which who knew anything about ML, but all were very nice. One guy is getting his masters program in AI, and asked me a question about using images in our search models. Adding images didn't work in our models because most information was already incorporated indirectly by the text and product hierarchy, which I tried to explain why it didn't work. I think he was just fixated on the fact that it didn't work. Another guy was fixated on image embeddings, and I explained why it was not necessary for what we were doing. Again, he didn't like the answer. I guess they really, really really want images in every model.
I was rejected for not using modern techniques. I appreciate the feedback, but this is what happens when you have conversations with people who don't have real-world experience in the model building process. They latch onto bits and pieces of what you have to say and fixate on the correctness of ideas versus having a real conversation about it.
Interview questions [1]
Question 1
Pytorch code question
Behavioral question
Past experience