Facebook - MLE onsite Interview feedback/recommendation/rant
I currently have 13 yoe in software and I am at a very senior role at my current company. (senior TL / ML / Architect). I lead a team of 10 people (Engineers, Data Scientists, and etc...). I was promoted recently to an Architect to lead multiple ML teams of the same size with more scope, responsibility and owing the product. company: #1 index fund company current tc: 200K I been preparing on leetcode for about 2-3 months now. I solved ~400 problems and solved the top 200 facebook tagged problems by frequency. I can do 1 mediums in about 20 mins. I also practiced 5-6 mock interviews on Pramp. Context: I had a phone interview last year(2020) and didn't pass. Recruiter reached out this year and shared the feed back this year. The feedback was that I did not provide the optimal solution. All other aspects were good, good communication and walking through the solution, solving without hints and etc. The problem was that I used recursion instead of iteration, which wasn't optimal space complexity. I didn't realize that this was even a concern as I solved the initial coding problem in haskell in about 2 mins and had to rewrite it in c# for the next 15 mins. I decided to copy my solution instead of just writing it from scratch. Anyways I decided to give another shot this year(2021). (Haskell has TCO, so the space complexity issue doesn't exist in my original solution). The interview was very stoic and didn't really help at all. Phone Round: 1. Coding Round - solved only 1 problem (lc easy). I provided multiple solutions. First answer was bad in space and time complexity. Second answer was optimal, but was more complicated than needed. The interviewer said he had to give me a hint to make it more simple. Anyways, the Interviewer advanced me to the first round as the rest of the interview was great. I proposed multiple solutions, provided optimal solution, walked through space and time complexity, examples, and discussed input constraints in length and accounted for all edge cases. Proceeded to the on-site interview. 1. Behavioral - It went very well 2. ML - It went exceptionally well 3. System design - went ok. They asked me something completely random not related to ML or stuff on my resume and was about infrastructure operations. Post interview, I reviewed their engineering blog and it was basically the same answer as the one I provided during the interview. So not bad. 4. Coding - bombed it. The question was related to min distance between two nodes of a tree. I provided two solutions, using parent pointers and running DFS with post order traversal. I evaluated the solutions and showed time and space complexity. I wanted to do the parent pointer solution was kinda forced to do DFS. He asked me why I didn't want to do it and I couldn't honestly answer that its going to take a long time since there are a lot of edge cases. I DID NOT FINISH coding the solution in remaining time as there was a lot of edge cases as he pushed me in this direction. Also, I seemed to have missed a very critical detail. In hindsight, I think the tree as actually a BST (binary search tree), which was a LC easy. The interviewer didn't really highlight this point at all and let me struggle for the rest of the interview. The interviewer was also very stoic like the interviewer I had during my first phone screen. Are they not supposed to give hints if you are on the wrong path? 5. Coding - Completely bounced back. Solved 2 lc med problems. Did an awesome job imo! Proposed multiple solutions, provided optimal solution, walked through space and time complexity, examples, and discussed input constraints in length. Interviewer was very engaged. I was even very candid saying I have seen 1 of the problem before. The Interviewer decided to give a variant that I have not seen. So, here is the feedback from the recruiter: 1. Behavioral - very good 2. ML - very good 3. System Design - good 4. Coding - not very good The recruiter did not provide me with any specific details. The recruiter was considering me for an E5 but asked me if I should be down-leveled to E4 since the feedback wasn't great. I said that I wanted to be an E5. I actually was expecting to be E6 compared to what I am doing now! The recruiter has a strong inclination that I will most likely be rejected. Anyways, I am not sure whats going to happen. Recruiter said if I have other opportunities, I should be pursuing those instead. The recruiter will submit my application to the quorum for judgement. Should I just retry again next year if I don't get the job? it seems that the problem was that I can't do the coding problems under pressure. None of the question I got were hard, but interview pressure is real. Is the recruiter just trying to low ball me? If they come back with an E4 offer, should I reject it? Recruiter said that this is a senior engineer. Looking at levels.fyi - this role is for about 3-6 yoe(junior). I would be completely over qualified for this role. I have mentored several junior developers and this would be embarrassing or maybe its just my hubris talking. Would love to know if I have a chance still? #facebook #tech
Levels.fyi
Fixing on a single company is wrong. Interview else-where and switch now. FB will be around next year if you want to interview again at that time.
I do have other prospects so that’s always good news.
E5 here is a very thankless job.
A real ML expert wouldn't waste his time doing leetcode and preparing such interviews.
They would if they wanted to work at Facebook and other places where Ml is held within the Engineering side. But I do agree ML is different.
You lead a team of 10 but the TC is $200k only?! 😬
I think the TC for a data science or ML manager is at $350k if not higher.
For 4, I think some interviewers are actually looking to see if you are asking the right questions to clarify the problem or not.
I would have felt better if they flagged me as inattentive than not mention that the tree is a binary search tree. The interviewer only went over in detail that tree had a left and right child. Nothing more. He might have given a BST as an example and might have have used the word "Binary Search Tree" once (which I didn't catch)
They are clearly lowballing you !!! I have about 8 years of experience and was considered for E4/E5 [No Offer] .I am very sure that you seem to have lot more experience in ML space and Software engineering. Given that you have done your preparation. Try interviewing elsewhere where you will be valued more and will be working on interesting ML work. Perhaps try other financial firms where your experience is valid ,Blackstone etc.
Why the hell would you even want to be E5 If you have 13 yoe?
TC for E5 is still better than what I make now.
I’d still not do it . You want something with better scope for your experience .