AI, data & engineering
AI / LLM Engineer
AI / LLM Engineer is an engineer who builds products on existing LLMs: connects models and RAG, measures answer quality and cost, and ships it to production.
Also known as: AI Engineer, LLM Engineer, Applied AI Engineer, GenAI Engineer, Generative AI Engineer, LLM Developer.
What does an AI / LLM Engineer do?
An AI / LLM Engineer is an engineer who builds products on existing models and owns how they behave for users. It is now the most common AI job title: most companies don't train models from scratch, they build existing ones into the product. Typical tasks:
- connects LLMs through APIs or deploys open-source models;
- builds retrieval (RAG): search over the company's documents and data that the model uses in its answers;
- writes and versions prompts and call chains;
- measures answer quality and cost and keeps track of them after launch;
- takes LLM features to production and investigates what breaks for real users.
When does a company need an AI / LLM Engineer?
A company needs an AI / LLM Engineer when it builds LLMs into its product or internal tools: an assistant, knowledge-base search, document processing, text generation. It is often the first AI hire in a product company, because most of these tasks are solved with existing models. The role doesn't fit when the task requires training your own models on your own data (scoring, recommendations, computer vision): then you need an ML Engineer.
AI / LLM Engineer vs ML Engineer: what's the difference?
An AI / LLM Engineer differs from an ML Engineer by working with existing models and owning their behaviour in the product, while an ML Engineer trains and fine-tunes models and builds ML pipelines. An Agentic AI Engineer is a special case of the AI / LLM Engineer who builds agents for multi-step tasks. The work prompt engineers used to do is now also mostly done by AI / LLM Engineers.
AI / LLM Engineer vs related roles: comparison
| AI / LLM Engineer | ML Engineer | Agentic AI Engineer | |
|---|---|---|---|
| Responsible for | How existing models behave in the product | Model training and ML pipelines | Agents that carry out tasks on their own |
| Main result | An LLM feature in production | A model of the required quality | A multi-step task done reliably |
| Works with | Product managers, backend team | Data engineers, data scientists | AI team, owners of internal systems |
Why is it hard to hire an AI / LLM Engineer?
It is hard to hire an AI / LLM Engineer because the title is new and widespread, while real experience with LLMs in production is rare.
- The title is used very broadly: engineers with years of LLM work and developers fresh from one course both use it, so a title search returns a lot of noise.
- The role needs strong backend engineering together with the ability to measure model answer quality; candidates usually have one or the other.
- Experience with quality evaluation and request costs barely shows in a CV.
- Engineers with LLM features in production are wanted by many companies at once and rarely look for work actively.
What to decide before hiring an AI / LLM Engineer?
- Which tasks does the model solve in the product, and how will you know an answer is good? Without a quality criterion the engineer will improve the feature at random.
- Existing models through APIs, or your own open-source models? Running models on your own GPUs is different experience and often a different candidate.
- Does the team have a quality evaluation setup? If there are no evals, this person will have to build them, and the profile should reflect that.
- What matters more for the role: backend engineering or ML experience? The answer decides which teams to search in.
Where to find an AI / LLM Engineer?
AI / LLM Engineers most often come from backend and full-stack engineers who have shipped LLM features to production, from ML engineers who moved into applied LLM products, and from AI startups and product teams with AI assistants. The signal in a profile is an LLM feature serving real users with a clear quality metric, plus open-source projects around RAG or model evaluation. NEWHR finds such candidates by the results of their work, not by applications to a job post.
How to assess an AI / LLM Engineer in an interview?
An AI / LLM Engineer is assessed by a walkthrough of an LLM feature the candidate shipped (architecture, how quality was measured, cost per request, what broke for users) and a work sample: a small RAG or LLM feature on your data. On the company side, the Head of AI or CTO takes part, along with the team's lead engineer.
Strong signals:
- explains how they measured quality and on how many examples;
- knows the cost per request and what they did to bring it down;
- explains when an LLM isn't needed and plain code will do.
Red flags:
- checked quality by eye on a handful of examples;
- all of their experience is demos and tutorials;
- can't explain why the model got specific cases wrong.
How long does it take to hire an AI / LLM Engineer?
We present the first candidates for the AI / LLM Engineer role within 5–7 days of the brief. The whole search depends on how rare the profile is: a senior ML team of six took us five months.
Related roles
- ML Engineer — trains and fine-tunes models rather than building existing ones into the product.
- Agentic AI Engineer — builds agents that carry out multi-step tasks on their own.
- Evals Engineer — owns the quality evaluation system the AI / LLM Engineer relies on.
- AI Infrastructure / Inference Engineer — makes models run fast and cheaply on your own GPUs.
- MLOps / LLMOps Engineer — owns deployment, versions and monitoring of models.