The process took 5 weeks.

Probably about 5 weeks. A quick recruiter chat, then a couple of technical phone screens covering standard coding and ML theory to make sure I actually knew my core algorithms. The onsite loop was four rounds covering DSA, ML system design, an ML domain deep dive, and their standard Googleyness behavioral chat.

The system design round is easily the biggest filter imo because you can't just talk about training a model; you actually have to architect an end-to-end production pipeline and figure out serving infrastructure at scale. The domain round gets deep into the weeds of loss functions and model tradeoffs, while the DSA rounds are just looking for clean code and quick edge case detection. If you are prepping, definitely shift your focus from model tuning to production-level ML.

I did a mock interview on Prepfully with a Google MLE beforehand, which completely saved my system design structure and gave me a solid reality check. Not an easy process, but honestly very fair.

What is the time complexity of training a support vector machine?