I work in the software industry and serve on committees/boards for several schools with which I have various affiliations.
I speak with all types of employers for all types of students, in my capacity as a member of those committees/boards. Ranging from high school dropouts finishing their high school diploma later in life to masters students. My comments here are mostly about traditional undergraduate students, though.
I’ve always been aware that there are (at least) two tiers of software engineering jobs but I have a hard time articulating the reasons for that. My high school friends seem to do roughly similar work to my FAANG-ish job, but for a fraction of the pay, and it bugs me because I know they are smart and deserve more.
A few thoughts follow.
With respect to software work, I have both personal experience and mentoring experience.
First, a lot of the "type of work at FAANG vs. small shop" is extrapolation and guesswork for me. I have experience at both types of places, but in extremely different role types. I've mostly only been in "normal" roles at low-prestige firms, but I've only been in either pure R&D groups or "elite" product groups at high-prestige firms. So, for me, FAANG-style work looks incredibly different and is completely inaccessible to the average student from my alma mater at basically any point in their career, without substantial additional education. A lot of this personal anecdote is due to comparing my career before and after my (more elite-signalling) PhD. So, it's hard for me to know.
Now, in terms of mentoring, I can say some things.
First, the higher paying firms consistently get our best and brightest students. These students are the ones who only pass through our type of institution once every 5-10 years. Those employees are way more versatile, productive, disciplined, and frankly just intelligent. Are they doing to same work? Sometimes yes, sometimes no. But FAANG/high finance/et al. does seem to be choosing the absolute cream of the crop. At least from us. They are definitely buying more versatility and higher general intelligence if nothing else. Again, FROM US, a small non-elite school. I know institutional bias means that this may not be true in general.
Second, my impression is that a lot of our students would have a really hard time simply onboarding to tooling at the two FAANGs I know the most about. Think "Java programmer who uses IntelliJ and drops a JAR onto a single VM. Knows about but still isn't very comfortable with things like command lines or version control. Relies on senior engineers for anything more complicated than developing on one or two branches with rare merges and deployment to a single VM". That's a bit of an exaggeration but not far off, so you can see how landing at Google or Facebook might be a painful way for those students to start their career.
FWIW, I think the department is very pragmatics-focused. We aren't failing to teach git because we're too busy teaching automata theory, for example. We don't even teach the latter. So, very focused on software development career prep. But the quality of student and dedication to craft just isn't there. Getting students into internships gives a huge improvement but is still an uphill battle for us. Improving internship placement is therefore a major focus of the alumni's strategic efforts over the next five years. I suspect that with a couple years of full-time work and good mentorship all of our students would be ready for a normal SWE role at FAANG, but straight out of undergrad would fail at even the basics of a FAANG software job.
Personally I don’t think my mental “versatility” got too much exercise there, except maybe in rescuing production systems in emergencies. But presumably if you stick enough smart people together someone will eventually invent something novel and profitable, even if the median employee never does.
Given the cost of every minute of downtime at FAANG scale, the ability to rescue a production system a few hours faster than your competition is worth embarrassing multiples of your wage premium against non-FAANG engineers.