Skip to content

Research report ยท Live data

Programming Languages in AI Training

Of 284 Coding & Software Evaluation roles open on 17 Oct 2026, 69.4% name a programming language, Python most often (35.6%); where enough roles on two or more platforms state pay, median advertised rates run from $50/hr (Python) to $160/hr (Lean (theorem prover)).

Data window
20 Sep 2026 to 17 Oct 2026
28 days, UTC
Computed
18 Oct 2026, 20:20 UTC
Recomputed every day
Source
The public job boards of the AI training platforms this site tracks, read every day. Languages are read from role titles on every platform, and from the full descriptions of the roles whose job board publishes one.
Method
Version 1. Definitions
CC BY 4.0, how to cite

Every figure on this page as CSV and JSON, sent to your work email.

open Coding & Software Evaluation roles read
284
of them name a programming language
69%
of coding roles name Python, the most named language
36%
median advertised for roles naming Python
$50/hr

Key findings

  1. 1

    Python is the language named most often in open Coding & Software Evaluation roles: 101 of 284 (35.6%), in the title or the description, ahead of TypeScript (68); 23 different languages are named, and 38.0% of the roles name one in the title.

    Figure 1
  2. 2

    Where a description is published, 72.0% of 239 Coding & Software Evaluation roles name a language and 38.5% name two or more; Python is named in 392 roles read.

    Figure 2
  3. 3

    Among languages with at least 8 roles stating pay on 2 or more platforms, Lean (theorem prover) advertises the highest median ($160/hr) and Python the lowest ($50/hr), against $65/hr for all Coding & Software Evaluation roles stating pay.

    Figure 3
  4. 4

    Anyone AI names a language in 93.1% of its 29 open coding roles (most often Python); Invisible in 45.2% of its 42.

    Figure 4
  5. 5

    353 open roles outside Coding & Software Evaluation name a programming language, most in Law, Medicine & Finance (212, most often Python) and Math & STEM (85, most often Python).

    Figure 5
  6. 6

    The frameworks and tools named most often in Coding & Software Evaluation descriptions are Git or GitHub (45 of 239), REST or GraphQL APIs (42 of 239) and CI/CD (41 of 239).

    Figure 6
  7. 7

    30.6% of open Coding & Software Evaluation roles name no language anywhere they are read: 21 have a generic title and no published description, and 66 have a description that names none.

    Figure 7

Figure 1

Programming languages named in role titles

Which programming languages do AI training job titles ask for?

LanguageOpen roles naming itOf them, CodingShare of coding roles
Python
59
3813.4%
JavaScript
12
124.2%
Java
10
103.5%
TypeScript
10
103.5%
KotlinSmall sample
7
72.5%
RustSmall sample
7
72.5%
SQLSmall sample
7
62.1%
Lean (theorem prover)Small sample
5
00.0%
C++Small sample
4
41.4%
GoSmall sample
4
41.4%
SwiftSmall sample
4
41.4%
C#Small sample
3
31.1%
LuaSmall sample
3
31.1%
Objective-CSmall sample
3
31.1%
PHPSmall sample
3
31.1%
HTML and CSSSmall sample
2
20.7%
Bash and shellSmall sample
2
20.7%
CSmall sample
1
10.4%
DartSmall sample
1
10.4%
MATLABSmall sample
1
00.0%
RSmall sample
1
10.4%
RubySmall sample
1
10.4%
ScalaSmall sample
1
10.4%

Sample. 135 of the 3,668 roles open at the end of 17 Oct 2026 name a programming language in their title, 108 of them among the 284 Coding & Software Evaluation roles.

Names that are also ordinary words or letters (Go, C, R, Swift, Rust, Ruby, Dart, Julia, Lean) count only in a programming context: an explicit phrase ('Golang', 'C/C++', 'R programming', 'Lean 4'), a list beside another language ('Java, Go, Rust'), or the head of a title before a role word ('Go Code Review Developer').

A title naming two languages counts for both. Share of coding roles: of all 284 open Coding & Software Evaluation roles. The median needs 8 roles stating an hourly rate; rows under 10 roles are marked.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 2

Programming languages named in role descriptions

Which languages do the full job descriptions name, in coding work and beyond?

LanguageRoles naming itCoding roles naming itShare of coding roles read
Python
392
9640.2%
TypeScript
67
6426.8%
Java
66
6125.5%
Go
55
5020.9%
JavaScript
55
4920.5%
SQL
55
2610.9%
MATLAB
51
31.3%
C++
48
2811.7%
R
39
31.3%
C#
37
3715.5%
Rust
21
177.1%
C
20
52.1%
Bash and shell
16
31.3%
Julia
11
00.0%
Kotlin
10
104.2%
SwiftSmall sample
8
83.3%
Coq (Rocq)Small sample
7
00.0%
HTML and CSSSmall sample
7
41.7%
FortranSmall sample
6
00.0%
Verilog and VHDLSmall sample
6
00.0%
PHPSmall sample
6
62.5%
VBA (Visual Basic)Small sample
5
10.4%
Lean (theorem prover)Small sample
4
00.0%
Objective-CSmall sample
4
41.7%
RubySmall sample
4
41.7%
DartSmall sample
3
31.3%
AgdaSmall sample
1
00.0%
HaskellSmall sample
1
00.0%
LuaSmall sample
1
10.4%
ScalaSmall sample
1
10.4%

Sample. 3,260 roles open on 17 Oct 2026 whose job board publishes a full description (AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI)), 88.9% of the 3,668 open roles; 239 of them are Coding & Software Evaluation roles.

Names that are also ordinary words or letters (Go, C, R, Swift, Rust, Ruby, Dart, Julia, Lean) count only in a programming context: an explicit phrase ('Golang', 'C/C++', 'R programming', 'Lean 4'), a list beside another language ('Java, Go, Rust'), or the head of a title before a role word ('Go Code Review Developer').

Descriptions often list several accepted languages ('Python, Java, Rust, Go, C++ or TypeScript'): each one counts, so a role is counted once per language it names.

The four platforms that publish descriptions are expert networks; their roles lean towards expert and senior work.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 3

Advertised pay by programming language

What does AI training work in each programming language pay?

LanguageMedian per hourMiddle halfRoles stating payPlatforms stating pay
Lean (theorem prover)
$160
$126 to $18584
MATLAB
$76
$41 to $96428
C
$75
$33 to $9095
C++
$75
$65 to $87349
R
$75
$60 to $113358
Rust
$75
$65 to $75228
Bash and shell
$65
$46 to $68107
C#
$63
$63 to $63355
Go
$63
$63 to $754911
Java
$63
$59 to $666011
JavaScript
$63
$60 to $705211
TypeScript
$63
$63 to $756411
SQL
$60
$43 to $784411
Kotlin
$58
$36 to $69104
Python
$50
$40 to $8034918
AgdaSmall sample
n/a
n/a11
Coq (Rocq)Small sample
n/a
n/a64
DartSmall sample
n/a
n/a33
FortranSmall sample
n/a
n/a31
HaskellSmall sample
n/a
n/a11
Verilog and VHDLSmall sample
n/a
n/a52
HTML and CSSSmall sample
n/a
n/a64
JuliaSmall sample
n/a
n/a11
LuaSmall sample
n/a
n/a32
Objective-CSmall sample
n/a
n/a31
PHPSmall sample
n/a
n/a43
RubySmall sample
n/a
n/a33
ScalaSmall sample
n/a
n/a11
SwiftSmall sample
n/a
n/a74
VBA (Visual Basic)Small sample
n/a
n/a53

Sample. Roles open on 17 Oct 2026 that name the language and state an hourly rate, of any kind of work; 257 of the 284 Coding & Software Evaluation roles state one (median $65/hr).

A language counts when the title names it or, for roles at AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI) (the job boards that publish a full description), when the description does.

Pay is the hourly figure the posting advertises (the middle of a stated range). Medians need 8 roles stating a rate; languages with fewer are listed last, without a median. The middle half runs from the 25th to the 75th percentile.

A median whose priced roles all come from one platform is marked and not ranked: it describes that platform's listings (often a family of roles at one advertised range), not the language.

Mostly in: the kind of work most of the roles naming the language belong to. A language named mainly outside coding (R in statistics roles, for example) is paid as that work is.

Roles that list several accepted languages carry one rate for all of them, so languages named together share their pay: the differences between them are smaller than the table's precision suggests.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 4

Languages named by each platform's coding roles

Which platforms hire coders for which languages?

PlatformOpen coding rolesNaming a languagePythonTypeScriptJavaJavaScriptGo
SuperAnnotate (SME)6687.9%22.7%16.7%10.6%6.1%3.0%
Alignerr4355.8%11.6%9.3%2.3%11.6%2.3%
Invisible4245.2%14.3%9.5%11.9%11.9%2.4%
micro13259.4%46.9%31.3%31.3%9.4%28.1%
Anyone AI2993.1%93.1%89.7%89.7%89.7%89.7%
Turing2286.4%68.2%27.3%22.7%27.3%22.7%
Mercor1656.3%25.0%12.5%12.5%0.0%12.5%
AfterQuerySmall sample944.4%11.1%0.0%11.1%11.1%0.0%
RemoExpertsSmall sample6100.0%66.7%33.3%33.3%0.0%50.0%
OpenTrain AISmall sample580.0%40.0%0.0%0.0%0.0%0.0%
Handshake AISmall sample425.0%25.0%0.0%25.0%25.0%25.0%
BraintrustSmall sample2100.0%100.0%0.0%0.0%0.0%0.0%
ClaruSmall sample2100.0%50.0%50.0%50.0%50.0%0.0%
TeracSmall sample20.0%0.0%0.0%0.0%0.0%0.0%
VettoSmall sample2100.0%100.0%50.0%50.0%100.0%50.0%
Pasiflora AISmall sample1100.0%100.0%100.0%0.0%100.0%0.0%
Sovrano AISmall sample10.0%0.0%0.0%0.0%0.0%0.0%

Sample. 284 Coding & Software Evaluation roles open on 17 Oct 2026, by the platform that lists them. Language columns: the share of the platform's coding roles naming each of the 5 languages named most often.

A language counts when the title names it or, for roles at AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI) (the job boards that publish a full description), when the description does.

A platform whose titles are generic ('Software Engineer') and whose board publishes no description shows few languages: that is what its listings say, not what its projects use.

Listings only: a count of open roles, not of hires or applications.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 5

Programming languages beyond coding roles

Which other kinds of AI training work ask for a programming language?

Kind of workRoles naming a languageShare of its open rolesMost named
Law, Medicine & Finance
212
19.1%Python
Coding & Software Evaluation
197
69.4%Python
Math & STEM
85
39.0%Python
AI Evaluation & Rating (RLHF)
27
18.9%Python
Agentic Tasks & RL Environments
14
21.9%Python
Data Annotation & Labeling
6
4.3%Python
Red-Teaming & AI Safety
6
11.1%Python
Writing & Content
2
5.9%Python
Robotics & Physical-AI Data
1
1.5%Python

Sample. 3,668 roles open on 17 Oct 2026; 353 roles outside Coding & Software Evaluation name a programming language.

A language counts when the title names it or, for roles at AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI) (the job boards that publish a full description), when the description does.

Shares are of each kind's open roles on every platform, including those whose descriptions are not published, so they are floors.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 6

Frameworks and tools named in coding role descriptions

Beyond the language, which frameworks and tools do coding roles name?

Framework or toolCoding roles naming itShare of coding roles read
Git or GitHub45
18.8%
REST or GraphQL APIs42
17.6%
CI/CD41
17.2%
Docker22
9.2%
Node.js19
7.9%
Databases (PostgreSQL, MySQL, MongoDB)18
7.5%
React18
7.5%
AWS14
5.9%
SwiftUI, UIKit or Jetpack13
5.4%
pandas or NumPy11
4.6%
IDEs (VS Code, IntelliJ, Xcode)10
4.2%
Next.js8
3.3%
Django, Flask or FastAPI8
3.3%
Azure7
2.9%
Google Cloud7
2.9%
CUDA and GPU kernels (Triton, NKI)7
2.9%
Kubernetes7
2.9%
Linux7
2.9%
Angular6
2.5%
AI coding assistants (Cursor, Copilot, Claude Code)5
2.1%
PyTorch5
2.1%
scikit-learn5
2.1%
Vue5
2.1%
Flutter4
1.7%
LangChain or LlamaIndex4
1.7%
React Native4
1.7%
TensorFlow4
1.7%
Spark, Kafka or Airflow3
1.3%
Game engines (Unity, Unreal, Godot)3
1.3%
Jupyter3
1.3%
Spring3
1.3%
Terraform or infrastructure as code3
1.3%
.NET2
0.8%
JAX2
0.8%
BI tools (Tableau, Power BI, Looker)1
0.4%
Hugging Face1
0.4%
LLM serving (vLLM, TensorRT)1
0.4%
Ruby on Rails1
0.4%

Sample. 3,260 roles open on 17 Oct 2026 whose job board publishes a full description (AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI)), 88.9% of the 3,668 open roles; 239 of them are Coding & Software Evaluation roles.

Words that are also ordinary words or other sciences' terms ('angular momentum', a laboratory flask, Unity) count only in a software context.

A tool named in a list of options counts like a requirement: the figure counts mentions.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

Figure 7

Coding roles that name no language

How many coding roles say which language the work is in?

What the posting saysOpen coding rolesShare
The title names a language108
38.0%
Only the description names one89
31.3%
The description names none66
23.2%
The title names none; no description published21
7.4%

Sample. 284 Coding & Software Evaluation roles open on 17 Oct 2026; descriptions are read for the 239 of them at AfterQuery, Anyone AI, Appen, Babel Audio, Braintrust, Claru, Handshake AI, Invisible, Mercor, micro1, Mindrift, OneForma, OpenTrain AI, Outlier (Scale AI), Pasiflora AI, Perle Labs, Pila8, RemoExperts, Remotasks, Sigma AI, Sovrano AI, SuperAnnotate (SME), Terac, TransPerfect DataForce, Turing, Vetto, Welo Data and xAI (SpaceXAI).

Names that are also ordinary words or letters (Go, C, R, Swift, Rust, Ruby, Dart, Julia, Lean) count only in a programming context: an explicit phrase ('Golang', 'C/C++', 'R programming', 'Lean 4'), a list beside another language ('Java, Go, Rust'), or the head of a title before a role word ('Go Code Review Developer').

The kind of work also holds some engineering roles outside software (CAD, hardware, audio engineering) where no programming language is expected: part of the roles naming none are those.

The median needs 8 roles stating an hourly rate.

Source. AITraining.jobs Research, Programming Languages in AI Training, live data.

For labs, data vendors and researchers

Buy the data behind this report

The figures above are a summary. Behind them are records we collect ourselves, every day:

  • 10,289 roles tracked, each with its history

    Every role on 29 job boards, read daily since 13 Jul 2026: its pay and place, and the first and last day it was open.

  • 6,200 apply clicks in 28 days

    Which roles people actually apply to, by field, pay and kind of work. It happens on this site, so nobody else can count it.

  • 39 lab and vendor hiring boards

    The job boards of 25 AI labs and 14 human-data vendors, read every day: who is hiring for human data, and who is hiring to buy it.

What you can buy

  • The records

    The rows behind any figure, with their history, not only the totals.

  • A regular feed

    The same figures every week or month, as files.

  • A custom cut or benchmark

    Your countries, languages or fields, or your own platform against the market.

Method and definitions

The public job boards of the AI training platforms this site tracks, read every day. Languages are read from role titles on every platform, and from the full descriptions of the roles whose job board publishes one. Every figure is computed by one database query over the window shown, the same query for the live report and for every edition. Each figure states its sample beneath it.

Language named
A role names a language when its title does, or, at the platforms that publish a full description, when its description does. Matching is on whole words, case-insensitive, with the documented pattern of each language. A role naming several languages counts for each; descriptions often list several accepted languages.
Names that are also words
Go, C, R, Swift, Rust, Ruby, Dart, Julia and Lean count only in a programming context: an explicit phrase ('Golang', 'C/C++', 'R programming', 'Lean 4', 'iOS/Swift'), a list beside another language ('Java, Go, Rust', 'Python or R'), or the head of a title before a role word ('Go Code Review Developer'). The bare name counts only written with its capital, so the verb 'go' and 'lean manufacturing' never do.
Coding role
A role this site classifies as Coding & Software Evaluation, one of the twelve kinds of AI training work. Languages named in other kinds of work (Math & STEM, AI evaluation, agentic tasks) are counted separately.
Pay by language
The median hourly rate advertised by the roles naming a language, of any kind of work. A median needs 8 roles stating a rate, and is ranked only when those roles come from at least 2 platforms; a median resting on one platform is shown, marked, and not ranked.
Framework or tool
A library, framework, cloud, infrastructure or developer tool named in a description (React, PyTorch, Kubernetes, AWS, Git), each with a documented pattern; words shared with other fields ('angular momentum', a laboratory flask) count only in a software context.
Coverage
Titles are read on every platform; full descriptions only where the job board publishes them (Mercor, micro1, AfterQuery and Braintrust, about three in ten open roles). Each figure states which it reads.
Open role
A role listed by a given day and still in its platform's feed that day.
Kind of work
Each role is classified from its title and description into one of the twelve kinds of AI training work this site tracks, from AI evaluation and coding to multilingual and audio work. Every kind links to its own page.
Window
Live: the last 28 complete days, in UTC. Edition: one calendar month, in UTC.

What this data cannot tell you

  • Platforms that recruit without a public job board are not counted. Several large ones invite workers to projects directly.
  • A listing is not a hire. Some platforms keep evergreen roles open for months, and some post one listing per language or country.
  • Descriptions are read only for the platforms that publish them, which are expert networks: description figures lean towards expert and senior work, and coding roles elsewhere are read from their titles alone, so many name no language.
  • Roles listing several accepted languages carry one rate for all of them, so the pay of languages named together differs less than the table suggests; a mention may be a preference rather than a requirement.
  • The coding kind of work also holds some engineering roles outside software (CAD, hardware, audio engineering), which name no programming language.
  • A pattern can miss a language written in an unusual way; nothing is inferred beyond what the text says (an 'iOS' or 'Android' title names a platform, not a language).

How the job feeds are read and roles are classified: site methodology. How fresh, complete and consistent the data is, checked every day: data quality and coverage.

Editions

The live report covers the last 28 complete days and changes every day. Each complete month is also published as an edition, frozen two days after the month ends, so a figure you cite from an edition never moves.

  • Live data (this page)Last 28 complete days, recomputed daily

Cite and reuse

Free to quote, chart and republish under CC BY 4.0 with a link to this page. Live figures change daily: cite a monthly edition for numbers that will read the same later, or keep the access date.

Citation

AITraining.jobs (2026). Programming Languages in AI Training (live data, 20 Sep 2026 to 17 Oct 2026). https://www.aitraining.run/research/programming-languages-in-ai-training (accessed 18 Oct 2026)

BibTeX

@misc{aitrainingjobs_programming_languages_in_ai_training_live,
  author = {{AITraining.jobs}},
  title = {Programming Languages in AI Training (live data, 20 Sep 2026 to 17 Oct 2026)},
  year = {2026},
  howpublished = {\url{https://www.aitraining.run/research/programming-languages-in-ai-training}},
  note = {Accessed 18 Oct 2026},
}
Data: every figure as CSV (one line per value) and JSON, sent to your work email.Pay and open-roles data

More research

All 78 research reports