How we recognise an AI engineer on GitHub
We count someone as an AI engineer when their public work builds products on top of models rather than training them: LLM applications, retrieval (RAG), agents and tool calling.
In practice, their repositories, descriptions or bio carry at least two such signals: frameworks like LangChain, LlamaIndex or DSPy, vector stores like Qdrant, pgvector or FAISS, model APIs, evaluation and tracing tools like Langfuse or promptfoo. Training-side work (PyTorch, JAX, computer vision) is classified as AI research instead, and one profile can carry both roles.
The stack inside the role
Model APIs come and go; the retrieval and orchestration layer is where the role is decided. The technologies most present in their repositories:
| # | Technology | Share of AI engineers |
|---|---|---|
| 1 | AI agent | 48% |
| 2 | Python | 38% |
| 3 | RAG | 37% |
| 4 | LLM | 36% |
| 5 | TypeScript | 28% |
| 6 | Node.js | 13% |
| 7 | PyTorch | 12% |
| 8 | React | 8% |
| 9 | AWS | 8% |
| 10 | LLM inference | 7% |
Which countries have the most AI engineers?
Volume follows the large tech hubs, where most LLM products are built. The index column shows where AI product work weighs most among local developers.
| # | Country | Developers | Share of local developers | vs index |
|---|---|---|---|---|
| 1 | United States | 19,000+ | 32% | 1.4x |
| 2 | United Kingdom | 3,300+ | 25% | 1x |
| 3 | India | 2,900+ | 32% | 1.3x |
| 4 | France | 2,500+ | 22% | 0.9x |
| 5 | Germany | 2,000+ | 23% | 1x |
| 6 | China | 1,700+ | 33% | 1.4x |
| 7 | Canada | 1,500+ | 29% | 1.2x |
| 8 | Japan | 1,300+ | 29% | 1.2x |
| 9 | Brazil | 1,000+ | 20% | 0.8x |
| 10 | South Korea | 870+ | 24% | 1x |
Min. 1,000 indexed developers and 30 AI engineers per country; counts rounded down. Method: by-country study.
China stands out: 33% of its indexed developers, 1.4x the index average.
Cities
| # | City | Share of located AI engineers |
|---|---|---|
| 1 | San Francisco | 10% |
| 2 | London | 5% |
| 3 | New York | 5% |
| 4 | Seattle | 4% |
| 5 | Paris | 3% |
| 6 | Boston | 2% |
| 7 | Tokyo | 2% |
| 8 | Los Angeles | 2% |
Where the role overlaps with others
A profile can carry two roles. 19% of AI engineers also qualify for AI research: a spec asking for both describes a real, if narrower, profile.
| # | Also classified as | Share of AI engineers |
|---|---|---|
| 1 | Software engineering | 51% |
| 2 | AI research | 19% |
| 3 | Infrastructure, DevOps and SRE | 19% |
How reachable are they?
| Signal | AI engineers | Index average |
|---|---|---|
| Public email | 31% | 27% |
| Available for hire | 20% | 15% |
| Pushed code, last 90 days | 69% | 54% |
Ask for the evidence rather than the title. A repository that wires retrieval, evaluation and a model API together says more than "5 years of AI", a claim nobody can have made about LLM applications for long.
Who employs AI engineers today?
| # | Company |
|---|---|
| 1 | Microsoft |
| 2 | Red Hat |
| 3 | |
| 4 | NVIDIA |
| 5 | IBM |
| 6 | AWS |
| 7 | Amazon |
| 8 | Carnegie Mellon University |
What they build themselves
| # | Repository | Stars |
|---|---|---|
| 1 | obra/superpowers | ★ 281,184 |
| 2 | mattpocock/skills | ★ 277,760 |
| 3 | affaan-m/ECC | ★ 239,759 |
| 4 | getify/You-Dont-Know-JS | ★ 184,937 |
| 5 | msitarzewski/agency-agents | ★ 149,954 |
For the language side of the same search, see the guides to hiring Python developers and hiring TypeScript developers. To write the query by hand first, use the GitHub user search builder.
Frequently asked questions
How many AI engineers are on GitHub?
At least 78,000: that is how many StarHunt's index places in the role as of October 2026, counting only work visible in public repositories.
Is an AI engineer the same as a machine learning engineer?
Not in how the work shows on GitHub. AI engineers build applications on top of existing models: retrieval, agents, evaluation. Machine learning engineers and researchers train and serve the models themselves, with PyTorch, JAX or inference servers. StarHunt classifies the second group as AI research.
Where are AI engineers concentrated?
San Francisco is the largest hub among those we can place on a map, and China has the highest concentration relative to its size.
Describe the product you are building and get engineers who have shipped retrieval, agents or evaluation in public, ranked on the last twelve months.
Find AI engineers