The reality is that FAANG-type companies are the only companies large enough to pay these salaries and with enough hires to keep it anonymous.

These sites are also heavy on selection bias. If someone arrives at this site with a $150K AI salary, they're not going to go to the trouble of submitting it and getting it accepted. Instead, they're going to start updating their resume and browsing job sites while they seek these higher salaries.

Just some anecdata, I’m in ml not in faang and make more than the top end of the survey in the OP.

If I were to give some guidance, I’d suggest skipping on faang as they top out where they top out, and go downstream to smaller public companies who are growing and where you can deliver real value. Companies like snapchat doubled in 1 year, square doubled in 1 year, etc etc. If you’re already joining a trillion dollar company you need to enter as a leader to earn the top end.

You’ll never see these stats because 1) you don’t gain anything from sharing and 2) it’s easy to find who you are at smaller companies.

I have this issue talking to people on HN all the time. Fundamentally it boils down to this: The people making mid six digits or higher don't want to share exactly how much they make because there is very little upside, and people vote based on their own perspective of "what's true" which slants towards average.

The other thing that I don't like is the money-worship around here. I don't want someone else to feel bad that they're "only" making $120k a year. I want more people to help their communities and get involved with our democracies to help craft better legislation to protect ourselves from an ever more technological world. It really, really doesn't matter what type of car you drive.

But I've noticed something over the years: The people that go out of their way to give back never hurt for money. Doesn't work both ways (I know plenty of wealthy people that are self-centred) but if you give back it's a happier, more fulfilled life and as a side bonus you get a lot of intros with great upside.

I'm curious to know how you got into ML? Did you go the degree route? PhD? I'm guessing that's where the top end of the salary range goes to but would be interested to know your personal path and if you have any advice on how to start.

PhD. From what I found, the most successful people are the ones that marry the technical with the business and consistently deliver value. So it’s not talk/hype, but real value and competitive advantage and ability to communicate and effectuate change.

And how does one deliver value, in detail?

Interesting, as per my comment below, this is something I tried and failed to do well. Would you consider getting in touch for a brief chat? You can reach me on mr475 [at] cantab [dot] net - thanks!

Spot on

I don't know... Glassdoor's system seems to work pretty well for collecting this kind of data. Sure the number of reported salaries is maybe too small to say anything specific about individual companies, but in aggregate places like Glassdoor and LinkedIn could probably tell us with high certainty what the distribution looks like.

In a world where (A) compensation is typically mostly in the form of liquid equity grants and (B) compensation amounts change rapidly year-on-year, Glassdoor's system vastly understates compensation in the field.

But you have to recognize that this system is very atypical outside of FAANGs. Most large companies have very modest stock incentive programs.

> FAANGs

Do we just use FAANGs to mean every West coast software company? Because otherwise it is very typical of non-FAANGs as well.

Co-founder of Levels.fyi here - it's not so much that these sites solely support FAANG companies, it's just harder to get other people to add their salaries. We're working to encourage folks at other companies to contribute through dedicated company pages [1], etc. It's had a real impact and we've started to collect much more data across more companies / roles. We're also working to incentivize contributions more and I do think we'll be in a position to have a more representative sample over the next few months.

[1] https://www.levels.fyi/company/

If I can provide some guidance, there is no upside to share your data when you’re at a small company because it’s really easy to find who you are. If there is a way to obscure the company, for example, generalize to market cap or size ($1-3b mkt cap company with 2500-5000 employees) and bucket that way.

I appreciate the transparency of the data collection. I go to Levels for my first look over the Glassdoor’s of the world because I can actually trust the provenance of the data.

I like this site. It's also depressing. I see that new hires at my company are making substantially more than I was at that level, even accounting for inflation.

If you’re on the seller of labor side of the transaction, you should be happy labor is selling for more. Typically if new hires are getting paid more, then experience people will be getting paid more too.

This does likely require some action on your side to achieve though.

Typically, no-one will come and give you that 50% you are missing out on by not having pushed for an extra 4.1% every year for 10 years.

At my current company, we're experimenting with adding a market rate salary review component to our regular performance review cycle - we have a target salary percentile (based on years of experience, skillset, location, etc with aggregate market data from PayScale) for where we start new hires in a given role, and if for whatever reasons the market shifts upwards and leaves someone on staff below the new-hire percentile, we adjust upwards. We hope that regular pay increases, incentives, etc mean nobody's compensation ever falls lower than market rate after they're hired, but we also want to make sure we have a mechanism in place to correct if it does.

This is a feature we've been asked to add to our salary comparison tool (https://peerwyse.com). The request was to be able to create peer groups that you, as the manager, thought were appropriate peers for your report. Then you would be able to see the probability the report is in the different quintiles.

In the specific context the manager requesting this feature thought their reports were being underpaid and was looking for evidence that they could use to lobby for more.

Our company does market comparisons. It helps, but only for the severely under paid. They have a wide range that they consider acceptable.

Based on what I've seen, one probably has to switch companies.

In general that would be true. The problem for me is that it requires switching companies. I'm not able to do that since I naively did the jobs I was assigned. Now I'm a Neoxam and FileNet resource with zero prospects.

Seems that's the major flaw in self-reported salary aggregation sites: Because it's self-reported, you're generally going to get only one side of the bell curve: The side that people feel good about reporting. The FAANG employees, probably just the subset of them that are well paid. Who wants to go to a web site to report to the world about their utterly modest, average salary? How depressing! I wouldn't want to.

When I go to a site like levels.fyi looking to compare my comp, I'd have to believe that I'm not seeing "average" engineering salaries, but an average of the very best salaries at the very best companies. EDIT: Looks like you addressed my question in your comment already. Removing it!

I'm not convinced you're actually seeing the "very best salaries."

I've not posted my compensation at some firms because it'd be too easy to identify me. I'd say that levels.fyi often _underreports_ what levels of compensation are available to people of a given level.

We're trying to solve the salary sharing problem more generally with https://peerwyse.com by having you submit estimates of the salaries of your LinkedIn connections and then sharing back the aggregated view.

If you bring a few colleagues from the same locality / industry your estimates of mutual connections can be astonishingly insightful without anyone needing to break the taboo of revealing their own compensation information.

That's the goal of my project: PredictSalary (https://predictsalary.com)

The goal of my project is to predict the salary range from job opportunities (and Linkedin profiles in the future) using Deep Learning method (and ad-hoc rules). My plan is to cover countries outside US as well.

My thoughts exactly.