Say you build features on top of language models and wonder what to call yourself on a CV. More and more companies have picked a name for you. On October 6, 2026 "AI engineer" was on 484 open tech roles at 191 companies we track (distinct roles, one role posted in ten cities counts once). That counts titles like "AI Engineer", "Applied AI Engineer", "AI Platform Engineer" and "Software Engineer, AI".
It is also growing. Among companies we have tracked since before mid-May, these titles were 2.8% of the roles opened between early June and early August, and 3.5% of the roles opened since early September. A 23% jump in share in a few months.
So what is the job, according to the people hiring for it?
It is mostly building on models, not training them
The stack named in these roles tells you most of what you need to know:
| What the role names | Roles | Share |
|---|---|---|
| Python | 278 | 57% |
| LLMs (any form) | 152 | 31% |
| TypeScript | 108 | 22% |
| AWS | 102 | 21% |
| LangChain or LangGraph | 66 | 14% |
| PyTorch | 53 | 11% |
| RAG or retrieval | 45 | 9% |
TypeScript outnumbers PyTorch two to one. LangChain and LangGraph outnumber PyTorch too. That is the profile of someone who wires models into products: retrieval, orchestration, tool calls, the application around the model. If you want to train models, the title you are looking for is usually "research engineer" or "machine learning engineer", which we covered separately.
A few smaller signals worth noticing: 27 roles name MCP, and 27 name Claude.
The part almost nobody writes down
Here is the oddity. Evals, the work of measuring whether an AI feature actually works, are close to absent from these postings. We found two roles whose stack names evals. Two out of 484.
We do not think that means nobody does evals. More likely it is not yet something companies know how to put in a job post, or it gets filed under "testing" or "quality". But if you are good at it, that is worth saying out loud in an application, because the posts are not asking for it by name.
Who is hiring
| Company | AI engineer roles |
|---|---|
| OpenAI | 38 |
| Anthropic | 28 |
| Mistral AI | 21 |
| Databricks | 13 |
| Airwallex | 12 |
| Distyl AI | 12 |
| Binance | 10 |
| GitLab | 10 |
The labs lead the table, but they are not most of the market. By our count about 30% of these roles sit at companies whose product is AI. The other 70% are at companies whose main product is something else: payments, developer tools, exchanges, SaaS. 111 of the 191 companies have exactly one of these roles open.
Pay, level and where
285 roles publish a yearly US dollar band. The median floor is about $181,000 and the median ceiling $250,000.
61% are senior, 34% mid-level and 5% junior, roughly in line with the market. Half are hybrid (238), 134 are remote within a country or region, 105 onsite, and 7 are labelled remote from anywhere.
If you want this title
Get good at the application layer: retrieval, tool calling, orchestration, and the TypeScript or Python service around them. That is what the stacks ask for.
Do not wait for a lab. Most of these roles are at companies whose main product is not AI, and 111 of the companies have exactly one such role open.
If you can measure whether an AI feature works, say so. The posts rarely ask, which is exactly why it stands out.
You can search these roles on remoet.dev/jobs, and an agent can search them without an account.
How we counted
Titles are matched with a pattern, so a few "applied AI scientist" titles are in and some AI work under plain "Software Engineer" titles is out. The AI-company split is our judgement on each company's main product. Stacks are extracted from post text by a model. The growth figure compares the share of new roles in two windows at companies we tracked through both. All of it is October 6, 2026.
If you are an AI engineer and your job looks nothing like this list, we want to know what it does look like. Tell us in Discord.
