I applied online. The process took 2 weeks. I interviewed at Forterra in Oct 2024
No offer
Negative experience
Average interview
Application
I applied online. The process took 4 weeks. I interviewed at Forterra (Midtown New York) in Sep 2024
Interview
Initial phone screen, then a technical interview with a member of the perception team. Interview had a reasonable coding problem followed by some perception questions. Perception questions included both DL/NN topics as well as a traditional perception/classification. Overall I enjoyed the interview and appreciated the perception questions asked. Advanced to full loop stage, but due to some restructuring of the perception team this was cancelled. Very frustrating after having made arrangements to take the whole day interviewing with them.
Interview questions [1]
Question 1
Standard DL/NN questions about common architectures. One niche DL for CV architecture question that tripped me up. Finally a traditional perception problem.
Initial interview (~ 15min phone screen / Zoom chat) with HR. Some questions asked: 1) why are you looking for a new position outside of current company, 2) tell me about yourself / elevator pitch, and 3) what is your experience with LiDAR sensors and data.
Second interview with directly with hiring manager via Zoom (with video). This interview was ~60min and was comprised of two parts: 1) coding exercise (leetcode type coding question while screen sharing with the hiring manager) and 2) a conceptual computer vision question about how to create a classifier to classify points with a LiDAR dataset (3D point cloud).
I think there would have been a third interview with multiple team members had I done well enough on the second interview.
Interview questions [5]
Question 1
Why are you looking for a new position outside of current company?
A set S of n integers originally contains all of the numbers from 1 to n (and no other numbers). Due to corruption, exactly one of the numbers in the set S is duplicated onto another number in the set. The corrupted set has the integers 1 to n, with one number repeated and one missing number. Neither the original nor the corrupted sets are guaranteed to be in order. The original array is not given to us.
Write a function which takes an integer array of arbitrary length at runtime (the corrupted set) as input, and returns the duplicated number and the missing number as an array.
How would you create a classifier to classify a point P in a LiDAR dataset as either belonging to a "blob"-type object (trees, etc), a "plane"-type object (side of building, road, etc), or a "line"-type object (street sign, etc) using only x, y, z information of the point in question and the x, y, z coordinates of other points in a given neighborhood R.
