I receive an average of 1.5 recruiter messages per day. I have to choose whom to speak with, whom to skip, and whom to tell that “the salary range is a little low.”
You should always reply to recruiters, even when you are not looking for a job. At least tell them that the range is too low or that the conditions do not work for you. I would like to believe this helps move the market in a better direction. Imagine a recruiter hearing the same feedback three or four times in a row: the seed has been planted, and perhaps they will tell the hiring manager that something is wrong and the range needs to go up.
The market favors candidates, but in places it is still deeply confusing. Let us unpack it.
The AI engineering market is currently split into two parts, roughly 40/60. Because the field is so young, there is no consensus about an AI engineer’s responsibilities or even the exact skills the role requires.
I see two main categories:
- AI/ML Engineer. This person understands how neural networks work under the hood. They have trained and fine-tuned models and cleaned datasets. Today, relatively few companies need that work; most want to deploy an agent quickly through an API. I am not saying this is good, only that it is in demand. They came from the ML or data science world and had to learn how to build a product and write maintainable code. They are moving away from doing everything in Jupyter notebooks and increasingly work with backends and infrastructure. They used to be the person with the strongest ML and data science knowledge; now they are learning to become a conventional software engineer. They understand complex subjects deeply, but those subjects are overkill for many use cases. Sometimes that depth works against them and sends them further down the rabbit hole.
- LLM Developer / AI Engineer. This person may never have touched PyTorch, and usually does not need to. They probably moved from another engineering discipline, most often Python backend development and less often frontend. They know how to plan work in sprints, are comfortable with well-scoped tickets, and are learning to accept that systems built on LLMs cannot be 100% reliable. They have learned the main concepts: RAG, vector databases, the OpenAI API, and LLM frameworks such as LangChain, LlamaIndex, and LangGraph. They have read 👉my post about choosing between RAG and fine-tuning. They approach the field from the other direction, learning the ML concepts that help them build a usable product. They do not go especially deep into data science, may not know how to clean a dataset, and may not know what validation loss is. They are usually a stronger software engineer than profile one, although that can occasionally become a disadvantage in AI engineering, where everything is fragile, unclear, unstandardized, and extremely interesting.
If you identify with the first profile, congratulations: you cover almost the full spectrum of what an AI engineer is currently expected to know. The pool of roles you can consider is enormous, and you can command serious compensation — for example, here.
If you are closer to the second profile, congratulations as well. There are still plenty of vacancies. You do not cover the entire pool, but you have your own advantages. Apply broadly: there is a good chance PyTorch appears only in the job description, while the actual work consists of moving structured AI-generated JSON from one place to another.
At some point these two profiles meet. I suspect that convergence point will define the expertise expected from every AI engineer once the market stabilizes.
There is also a new emerging profile: the AI TypeScript Engineer. I wrote about that phenomenon here, and I expect the direction to keep growing.
I started as the first type. Today I describe myself as “a good balance between a deep understanding of neural networks and an engineering approach to integrating them into real products.”
AI engineering is currently extremely popular in the EU, so English is one of the main bottlenecks for Russian-speaking engineers. You need very strong English, and soft skills matter enormously if you are targeting salaries above €5–6k.
If you want to improve your English, Preply is a decent place to buy a few lessons and try different teachers. I found an excellent Chinese teacher there.
@makebugger
