If you're starting at the same Menlo Park company that I work at, for the long term, you may want to consider taking BART down to Fremont and then arranging a ride in a carpool/van-pool over the Dumbarton bridge: there are tons of folks who live in Fremont/Newark/Union City and who wouldn't mind a +1 for carpool lane. Many of them also have kids so they tend to come a bit earlier than others (most folks come in between 9:45 and 10:45).

Also, I've yet to have had a semi-recurring meeting schedule before 11 am (occasionally there are interviews earlier in the morning, however).

That said, irrespective of where in Bay Area you live, Ryan is 100% right: the problem isn't the average commute time (which is very easy to keep under 45 minutes -- this is not LA!), it's the outliers. It's a lot like the JVM: good performance on average but then a Concurrent Mark and Sweep failure happens...

Also, congratulations! I'm on the data/systems infrastructure as opposed to machine learning side (but folks move around quite a bit and a lot of teams do both ML and systems programming), but let me know if you've any questions.

80 (i.e. along the shore in Berkeley) is really bad quite often, but you could possibly skip most of that. If you can avoid that, it's not too bad, although the 880 parts in Oakland (due to port/construction traffic) can be bad.