Why the FDE salary median on Levels.fyi is higher than in job posts

Ask what a Forward Deployed Engineer earns in the US and you will get three answers. On 4 November 2026, Levels.fyi showed a median of $223,000, based on 155 salaries reported by users.

Fast AI Jobs, using November 2026 data, gives a median of $182,000. Recruiting From Scratch, in its June 2026 report, arrives at $195,000. Same job, same market, and the two ends are $41,000 apart.

If you are about to negotiate an offer or weighing a move from SWE to FDE, that gap affects you directly. None of the three numbers is wrong, but each measures something different. Reading FDE pay is therefore much like the daily work of an FDE: before trusting a number, ask where it came from.

What does $223,000 actually include?

Levels.fyi defines total compensation as base plus annualised stock plus bonus. So the $223,000 median cannot be set beside any figure that counts base alone. It is also self-reported data: users enter their own offers or earnings.

The distribution on the page shows more. P25 is $170,000, P75 is $290,000 and P90 is $351,400. P25 and P75 are $120,000 apart. One reasonable explanation is that stock varies widely between companies, pulling the two ends of the distribution apart.

The median does not stand still either. With self-reported data, each new batch of entries can move it. When you cite Levels.fyi, note the date you looked.

Job posts tell a different story

The other two sources do not ask workers; they read job posts. Fast AI Jobs takes the salary ranges in AI startup listings, reduces each range to its midpoint and aggregates across listings. The result is a median of $182,000 from 259 published salaries.

Recruiting From Scratch collected 924 FDE job posts from company careers pages. It states plainly that the data is based not on surveys but on job posts and real hiring cases. The median is $195,000, with P25 to P75 running from $160,000 to $215,000.

That P25–P75 band is only $55,000 wide, less than half of Levels.fyi’s $120,000. A narrower band is to be expected: one side measures ranges that companies publish themselves, the other adds stock and bonus as reported by each individual. The table below sets the sources side by side by method.

Source	Collection method	Pay included	Sample size	Median
Levels.fyi (US)	Self-reported by users	Base + annualised stock + bonus	155	$223,000
Recruiting From Scratch	Job posts on careers pages	Advertised salary range	924 posts	$195,000
Fast AI Jobs	AI startup listings, midpoint taken	Advertised salary range	259	$182,000
Levels.fyi (UK, FDSE title)	Self-reported by users	Base + stock + bonus	35	£139,361
Levels.fyi (India)	Self-reported by users	Base + stock + bonus	12	₹1,770,112
Salary.com (India)	Global Salary IQ data, April 2025	Not stated	Not stated	₹14.07 lakh

The UK row has one easily missed detail: the page sits under the FDSE (Forward Deployed Software Engineer) title, not FDE. Because the data is filed by job title, searching for only one name can mean missing part of it.

India: a median from 12 people

In the US, the two methods give two numbers, but at least each has a meaningful sample. India is different. Levels.fyi shows a median of ₹1,770,112, just over 17.7 lakh, but from only 12 reported salaries.

The spread across those 12 people is very wide. P75 is ₹4,223,893 and P90 is ₹5,592,066, meaning P75 is already more than twice the median.

With 12 data points, just a few people working for US companies with offices in India can lift the whole upper end. The page’s heading also reads United States, so you cannot be sure the location label is correct.

Salary.com comes in lower: a median of about ₹14.07 lakh, P25 of ₹11.29 lakh and P75 of ₹16.54 lakh. The data is dated April 2025 and gives no sample size. The Levels.fyi P75 is more than two and a half times the Salary.com P75.

Neither source is reliable enough to settle on a single number. The first has few data points; the second does not say how many it has. If you work as an FDE in a market with data this thin, the sensible approach is to treat both as a wide reference range and look for more job posts that list pay in that market itself.

One offer, two verdicts

Imagine you receive an FDE offer in the US of $200,000 base, with no stock and no bonus. Against the Levels.fyi median of $223,000, it is $23,000 short. Against the Recruiting From Scratch median of $195,000, it is $5,000 ahead. Same offer: one source says negotiate, the other says it is already good.

The mistake is comparing the wrong kind of number. This offer is all base, so setting it beside Levels.fyi total compensation, which includes stock, is not a fair test. Conversely, if the ranges in job posts are base only, which is worth asking the recruiter directly, then $200,000 sits inside the $160,000–$215,000 P25–P75 band, in the upper half.

Now try a different offer: $175,000 base plus $50,000 of stock a year. The total is $225,000, close to the Levels.fyi median. The base is below the $195,000 job-post median but still above the $160,000 P25, in the lower half of the band.

On paper this offer is more balanced, but the $50,000 in stock is only an annualised figure. If the shares vest over several years, you receive the full amount only if you stay long enough; if the company is not listed, the value rests on a valuation that can fall.

Ask about the vesting schedule and how the share price is calculated before treating this part as equal to cash.

The working rule is simple. Compare total compensation with self-reported data that includes stock and bonus. Compare base with advertised salary ranges, after asking exactly what those ranges include. Never compare a number of one kind with a number of the other.

Reading pay data is an FDE skill too

Who entered the data, when, which fields it covers, how many rows there are: these are the questions an FDE asks when handed a customer’s data. A dashboard showing revenue growth means nothing without knowing how revenue is defined. A salary median is no different.

So when negotiating, cite the source with its method rather than just a number. Saying “the self-reported total compensation median on Levels.fyi, 155 data points, checked on 4 November 2026” is far more persuasive than “I saw $223,000 online”.

Citing it this way shows the person across the table that you read data carefully, exactly the habit an FDE needs to bring to a customer.

When reading job descriptions, notice which posts state a clear salary range and which do not. When preparing for interviews, practise splitting an offer into base, stock and bonus the moment a recruiter reads out the number.

The FDE salary median will keep moving with each new batch of data. The number may change, but the habit of asking what it measures will last, in salary negotiations as much as in work with customers.