The tech lead described Telsa's infrastructure for AutoPilot. The entire ML platform is built in-house on-prem, not in the cloud. They have huge clusters. They may even consider building their own chip. They have a few senior engineers for developing the ML infrastructure.
Interview questions [1]
Question 1
System type questions:
- Module variable scope vs. function variable scope. Defined a variable in the module, and then defined a variable with the same name in a function, and asked what value would be seen in or outside of the function.
= How does Python deal with integer precision?
- How to determine the precision of a float in Python? Eg. 32 vs 64 bit? (Hint take diff and see how the diff is rounded)
