What the data shows

1
0 of the 11 groups shown (kinds of work, languages and professional fields) carry enough data on every ratio to get a rank; the others still appear, with a small-sample mark.
Figure 4

Figure 1

Scarcity by professional field

When a robot training job title names a profession or discipline, which of those jobs struggle most to find people?

Professional field	Scarcity index	Applicant shortfall	Slow closing	Pay premium	Open on 25 Oct 2026
All roles (the market)	1.00x	1.00x	1.00x	1.00x	1,139
Business and managementSmall sample	1.31x	1.33x	1.31x	1.29x	67
Engineering (non-software)Small sample	n/a	5.00x	n/a	n/a	37
Software and ITSmall sample	n/a	0.36x	1.00x	n/a	8
Finance and accountingSmall sample	n/a	5.00x	5.00x	n/a	1

Base: 4 professional fields that appear in the job titles of jobs still open on 25 Oct 2026. Apply clicks counted from 28 Sep 2026 to 25 Oct 2026; how long listings lasted, read over 159 jobs listed from 14 Sep 2026 to 25 Oct 2026.

Scarcity index: the three ratios to its right combined as a geometric mean. Each ratio is first clamped to the range 0.20 to 5.00, so a single extreme value cannot carry the index by itself. A typical robot training job scores 1.00, and a score over 1 marks work that is harder to fill.

Applicant shortfall: how many apply clicks this group's jobs should have received from 28 Sep 2026 to 25 Oct 2026 if they had drawn clicks at the market's rate for each open job, divided by the clicks they really received (the inverse of the demand index). A value over 1 means each opening drew fewer candidates than a typical job.

Slow closing: for the group's jobs listed from 14 Sep 2026 to 25 Oct 2026, the removals we would expect if every job had been taken down at the pace its own employer takes down jobs of that age (that employer's Kaplan-Meier hazard), divided by the removals observed up to 25 Oct 2026. A value over 1 means these jobs stay posted longer than the employer's other jobs. The comparison is made within each employer because posting habits, more than hiring speed, decide how long a listing stays up.

Pay premium: the median hourly rate advertised by the group's jobs still open on 25 Oct 2026, divided by the median of all open jobs that advertise a rate ($27/hr).

Rows get a rank only if the market's click rate predicted 0.6 or more apply clicks for its roles (a weight fitted on the clicks of every professional field in the table; under it, the table's pooled demand says more about the row than its own count does), enough of its listings came down (or stayed up) to set its pace apart from the other professional fields (in this table no row can: the gaps between their paces are within what chance alone produces), and 3 or more of the open roles behind them advertise an hourly rate. Rows that miss any of these carry a small-sample mark. A ratio sitting at the 5.00 cap means the group recorded no apply click, or no removal, at all.

A job joins a professional field row when its title contains the field as a whole word (the same word list the other research reports use); a title naming two professional fields is counted under both, and a title naming none is left out of this table. Most robot training titles name no professional field, so this table is often short or empty.

Credit: RobotTraining.jobs Research, Which Robot Training Roles Are Hardest to Staff, daily data.

Figure 2

Scarcity by language

For robot training jobs that call for a particular language, which languages are the hardest to hire for?

Language	Scarcity index	Applicant shortfall	Slow closing	Pay premium	Open on 25 Oct 2026
All roles (the market)	1.00x	1.00x	1.00x	1.00x	1,139

Base: 0 languages that appear in the job titles of jobs still open on 25 Oct 2026. Apply clicks counted from 28 Sep 2026 to 25 Oct 2026; how long listings lasted, read over 159 jobs listed from 14 Sep 2026 to 25 Oct 2026.

Pay premium: the median hourly rate advertised by the group's jobs still open on 25 Oct 2026, divided by the median of all open jobs that advertise a rate ($27/hr).

Rows get a rank only if the market's click rate predicted 1 or more apply clicks for its roles (a weight fitted on the clicks of every language in the table; under it, the table's pooled demand says more about the row than its own count does), the employers' own pace predicted 1 or more removals for its roles (a weight fitted on the removals of every language in the table; under it, the table's pooled pace says more about the row than its own count does), and 3 or more of the open roles behind them advertise an hourly rate. Rows that miss any of these carry a small-sample mark. A ratio sitting at the 5.00 cap means the group recorded no apply click, or no removal, at all.

A job joins a language row when its title contains the language as a whole word (the same word list the other research reports use); a title naming two languages is counted under both, and a title naming none is left out of this table.

RobotTraining.jobs lists the jobs open in a visitor's own country first, so jobs restricted to a handful of countries (as language jobs often are) collect fewer clicks for that reason alone.

Credit: RobotTraining.jobs Research, Which Robot Training Roles Are Hardest to Staff, daily data.

Figure 3

Scarcity by kind of work

Teleoperation, demonstration capture, test driving, drone flying: which kind of robot training work is hardest to fill?

Kind of work	Scarcity index	Applicant shortfall	Slow closing	Pay premium	Open on 25 Oct 2026
All roles (the market)	1.00x	1.00x	1.00x	1.00x	1,139
AV & Drone OperatorsSmall sample	1.92x	5.00x	1.21x	1.17x	528
Robot TeleoperationSmall sample	1.63x	5.00x	0.83x	1.04x	74
Robot Fleet OperationsSmall sample	1.19x	1.77x	0.96x	1.00x	254
Data & Demonstration CollectionSmall sample	0.92x	0.36x	1.96x	1.11x	124
Egocentric & Wearable CaptureSmall sample	0.50x	0.43x	0.54x	0.56x	139
Embodied-AI Data AnnotationSmall sample	0.39x	0.20x	0.81x	0.37x	12
Robot Task Evaluation & QASmall sample	n/a	0.72x	0.33x	n/a	8

Base: 7 kinds of work covering jobs still open on 25 Oct 2026. Apply clicks counted from 28 Sep 2026 to 25 Oct 2026; how long listings lasted, read over 159 jobs listed from 14 Sep 2026 to 25 Oct 2026.

Pay premium: the median hourly rate advertised by the group's jobs still open on 25 Oct 2026, divided by the median of all open jobs that advertise a rate ($27/hr).

Rows get a rank only if the market's click rate predicted 0.3 or more apply clicks for its roles (a weight fitted on the clicks of every kind of work in the table; under it, the table's pooled demand says more about the row than its own count does), enough of its listings came down (or stayed up) to set its pace apart from the other kinds of work (in this table no row can: the gaps between their paces are within what chance alone produces), and 3 or more of the open roles behind them advertise an hourly rate. Rows that miss any of these carry a small-sample mark. A ratio sitting at the 5.00 cap means the group recorded no apply click, or no removal, at all.

Each job sits in exactly one kind of work.

Credit: RobotTraining.jobs Research, Which Robot Training Roles Are Hardest to Staff, daily data.

Figure 4

Inside each scarcity score

Which clicks, listing lifetimes and advertised rates produce each group's score?

Group	Grouped by	Open on 25 Oct 2026	New listings	Weekly clicks per open job	Demand index	Median advertised pay	Median days listed	Removed	Expected from employer pace	Scarcity index
All roles	Market	1,139	160	0.09	1.00x	$27	38.5	24	24.0	1.00x
AV & Drone OperatorsSmall sample	Kind of work	528	81	0.02	0.17x	$32	n/a	6	7.3	1.92x
Robot TeleoperationSmall sample	Kind of work	74	8	0.02	0.19x	$28	6.6	3	2.5	1.63x
Robot Fleet OperationsSmall sample	Kind of work	254	53	0.05	0.56x	$27	n/a	3	2.9	1.19x
Data & Demonstration CollectionSmall sample	Kind of work	124	2	0.25	2.77x	$30	n/a	3	5.9	0.92x
Egocentric & Wearable CaptureSmall sample	Kind of work	139	9	0.21	2.34x	$15	12.0	5	2.7	0.50x
Embodied-AI Data AnnotationSmall sample	Kind of work	12	6	1.52	16.79x	$10	n/a	3	2.4	0.39x
Robot Task Evaluation & QASmall sample	Kind of work	8	1	0.13	1.39x	n/a	n/a	1	0.3	n/a
Business and managementSmall sample	Professional field	67	13	0.07	0.75x	$35	n/a	3	3.9	1.31x
Engineering (non-software)Small sample	Professional field	37	26	0.00	0.00x	n/a	n/a	0	0.0	n/a
Software and ITSmall sample	Professional field	8	3	0.25	2.80x	n/a	n/a	1	1.0	n/a
Finance and accountingSmall sample	Professional field	1	1	0.00	0.00x	n/a	n/a	0	0.5	n/a

Base: All groups from the three scarcity tables. Open jobs and advertised pay as of 25 Oct 2026; new listings and apply clicks from 28 Sep 2026 to 25 Oct 2026; listing lifetimes and removals across 159 jobs listed from 14 Sep 2026 to 25 Oct 2026, with each employer's back catalogue excluded.

Weekly clicks per open job: the apply clicks a group's jobs got on RobotTraining.jobs, divided by the days those jobs were open and multiplied by 7. Demand index: the group's rate divided by the market rate (0.09).

Median days listed: the Kaplan-Meier median time between a job's listing date and the last day it was seen, read the same way as in the listing-lifetime report. It shows n/a when over half the group's jobs are still up at the oldest age that can be measured, or when under 3 jobs were posted that far back. Because that median blends each employer's posting habits with the real speed of hiring, the index relies on removals compared with the same employer's pace instead.

Removed: jobs in the cohort that were taken down. Expected: how many of those same jobs would have come down at their employer's hazard for each age. Summed over the whole market, the two numbers match.

Rows get a rank only if it drew enough apply clicks to set its demand apart from the pooled demand of its table's other rows, by a weight fitted on those rows' own clicks, enough of its listings came down (or stayed up) to set its closing pace apart from the pooled pace of its table's other rows, by a weight fitted on those rows' own removals, and 3 or more of the open roles behind them advertise an hourly rate. Rows that miss any of these carry a small-sample mark. A ratio sitting at the 5.00 cap means the group recorded no apply click, or no removal, at all.

Median advertised pay: the hourly rate a listing publishes (the midpoint when it gives a range), shown only when 3 or more jobs publish one.

Credit: RobotTraining.jobs Research, Which Robot Training Roles Are Hardest to Staff, daily data.

Figure 5

Hardest to staff: above the market on all three ratios

Which robot training work lacks applicants, stays posted longer and pays more than the market, on all three counts together?

Nothing to show for this window.

Base: 0 of 0 ranked groups (kinds of work, languages and professional fields with enough data for all three ratios), as of 25 Oct 2026.

A group appears here only when each of its three ratios exceeds 1.00: each opening draws fewer candidates than the market, its jobs stay posted longer than the same employers' other jobs, and its median advertised rate beats the market median.

One job may belong to a kind of work, a language and a professional field together, so the same jobs can sit behind more than one row.

Credit: RobotTraining.jobs Research, Which Robot Training Roles Are Hardest to Staff, daily data.

Robotics labs · Fleet operators · Data vendors · Research teams

Buy the data behind this report

This report sums up records we gather ourselves, day after day:

1,317 robot training roles followed, history included
Every role on 85 employer job boards, read each day since 12 Sep 2026: its pay, its place, and the first and last day it was up.
451 apply clicks over 28 days
Which robot training roles people really apply to, by kind of work, pay and place. The clicks happen here, so no one else can count them.
19 sourced industry datasets
Funding, revenue, deals, fleets, deployments and prices across the robot training industry, each record linked to the publication it comes from.

On offer

The raw records
Every robot training role we track: title, employer, place, kind of work, pay as stated, and the days it was first and last open.
A regular feed
New and closed roles every week or month, as file drops or through the data endpoints.
A custom cut
Any figure here by country, city, kind of work, vehicle, employer or date, or your own hiring benchmarked against the market.

Need the people behind this data for your own robots? (you contract them yourself) or (we deliver the robot data).

How it was measured

Every day this site reads the public job boards of the robot training employers and programs it follows. Apply clicks come from visitors on RobotTraining.jobs; professions and languages are read from role titles. A single database query produces each figure for the window shown, and the live page and every monthly edition run that identical query. The note under each figure gives its base.

Staffing-difficulty index: The geometric mean of the applicant shortfall, the slow-closing ratio and the pay premium, each capped between 0.2 and 5 so no single extreme reading settles it. 1.00 is the typical robot training role; above 1, the group is harder to staff.
Applicant shortfall: How many apply clicks a group's roles would have drawn at the market's rate per open role, divided by the clicks they actually drew in the window: the inverse of the demand index. Above 1 means fewer applicants per opening than the market.
Slow closing: For each role posted in the last 26 weeks (none before 14 Sep 2026, back catalogues left out), its own employer's Kaplan-Meier curve gives the chance it comes down at each age; added over the group's roles, that is how many removals the group would see at its employers' pace. Slow closing divides those expected removals by the removals seen. Above 1, the group's roles stay up longer than others of the same age at the same employers.
Pay premium: The median posted hourly rate of the group's open roles over the median of every open role that gives one. It needs 3 or more roles with a stated rate.
Median days listed: The Kaplan-Meier median from listing date to last sighting, as in the listing-lifetime report. It sits next to the index but is not part of it, because it depends first on how each employer posts.
Ranked: A group enters the ranking when its clicks stand on their own against the demand of the groups it is set beside (the clicks its roles would have drawn at the market rate carry the weight of their pooled demand, fitted on their own clicks), its removals stand on their own against the closing pace of those groups (its expected removals carry the weight of their pooled pace, fitted on their own removals), and 3 or more of its open roles give an hourly rate. Every other group is shown with a small-sample tag.
Profession and language: Named in the title, matched as whole words against the profession and language vocabularies. A title naming two counts for each.
Open role: On a given day, a role that has been posted and still shows in the employer's feed.
Kind of work: A role's title and description place it in one of the seven kinds of robot training work this site follows, ranging from robot teleoperation and demonstration collection through egocentric capture, embodied-AI annotation, robot evaluation and fleet operations to AV and drone operation. Each kind has a page of its own.
Window: Live page: the 28 most recent full days (UTC), with listing lifetimes over every role posted in the 26 weeks to the same end, none before 14 Sep 2026. Edition: a single calendar month, with lifetimes over every role posted in the 26 weeks to month end, followed up to that end.

Limits of this data

Applicant interest is measured among RobotTraining.jobs visitors, not the whole market, and roles placed higher collect more clicks; visitors see the roles open in their own country first, which lowers clicks on roles tied to one site.
A listing coming down is not a hire: it may have filled, its program may have ended, or the employer may post it again. Setting it against the same employers removes their posting habits, not the reason one role closed.
These are posted rates, not earnings, and part of any gap between groups is where they hire.
Most titles name no profession or language, so those groups are small and their counts are floors; a role can count for a kind of work, a profession and a language at once.
Hiring that happens without a public job board is not captured.

Collection and classification of listings are explained on the methodology page. Freshness, completeness and consistency are tested daily in the data quality report.

Monthly editions

The live page spans the 28 most recent full days and updates daily. Every finished month also gets its own edition, locked two days after month end, so a number quoted from it stays fixed.

Live page (you are here)28 most recent full days, rebuilt daily

Quoting this report

You may quote, chart or republish these figures under CC BY 4.0, provided you link back to this page. The live numbers shift every day. For a figure that stays put, quote a monthly edition, or note the date you accessed this page.

Citation

RobotTraining.jobs (2026). Which Robot Training Roles Are Hardest to Staff (live data, 28 Sep 2026 to 25 Oct 2026). https://robottraining.jobs/research/hardest-robot-training-roles-to-staff (accessed 26 Oct 2026)

BibTeX

@misc{robottrainingjobs_hardest_robot_training_roles_to_staff_live,
  author = {{RobotTraining.jobs}},
  title = {Which Robot Training Roles Are Hardest to Staff (live data, 28 Sep 2026 to 25 Oct 2026)},
  year = {2026},
  howpublished = {\url{https://robottraining.jobs/research/hardest-robot-training-roles-to-staff}},
  note = {Accessed 26 Oct 2026},
}

Files: every figure in one CSV (a row per value) or as JSON, emailed to your work address.Other open data: datasets and API

Keep reading

Browse all 85 reports