The short answer
A strong LLM engineer job description says what product the person will build with models, what "good" means and how quality is measured, the production constraints (latency, cost, data sensitivity) and the stack. List four to six must-haves tied to shipped work, such as retrieval, evaluation and guardrails, rather than years of experience with a technology that is only a few years old. Then say how the interview process works.
Key takeaways
- Decide the problem, the quality bar and the constraints before you write. The job description is the brief made public.
- Lead with what the person will build in their first six months, not with a list of frameworks.
- Ask for evidence of shipped work: an LLM feature with real users, an evaluation suite, a production failure handled.
- "Years of LLM experience" requirements signal that the team doesn't know the field. Ask for outcomes instead.
- Many jurisdictions now require a pay range in the job ad or before the interview, including the EU's pay transparency rules and several US states.
- Gender-coded wording can make a role less appealing to qualified people. Keep the language plain and specific.
On this page6 sections
What to decide before you write
Write the job description last, after you've answered five questions. If you can't answer them, the ad will be vague, and vague ads attract keyword-matchers rather than people who have shipped.
- What will this person build in the first six months? "An assistant that answers customer questions from our help centre and order data" is a brief. "Drive our AI strategy" is not.
- How will you know it works? A user metric, an evaluation set, a human review process, or all three.
- What are the constraints? Response time, cost per request, data sensitivity, regulation, languages.
- What already exists? Data sources, models or providers in use, infrastructure, prior attempts.
- Is this really an LLM engineer? If the job is mostly training models on your own data, you're hiring an ML engineer. Our comparison of LLM engineer vs ML engineer vs AI engineer helps you decide.
The demand behind this role is real. LinkedIn's Jobs on the Rise 2026 ranked AI engineer as the fastest-growing job in the US, with retrieval-augmented generation (RAG), LangChain and PyTorch among the skills it associates with the role [1]. That popularity is also why so many ads for it look the same. The specifics are what make yours stand out.
The template
Copy this, then replace everything in square brackets. Delete any line you can't make specific.
Title
LLM Engineer (or Applied AI Engineer). Add the level if it matters, such as Senior. Avoid playful or invented titles: candidates don't search for them, and they say nothing about the work.
About the role
We're hiring an LLM engineer to build [product or feature] for [users]. In your first six months you'll [concrete outcome, e.g. ship an assistant that resolves common support questions using our help centre and order history]. You'll work with [team: product manager, backend engineers, designer] and report to [role]. We work [remote / hybrid / on-site], and you'll need to overlap with [time zone] for [number] hours a day.
What you'll do
- Design and build LLM features in our product: retrieval over [data], prompts and context, tool or function calling, and [agent workflows, if relevant].
- Build and maintain evaluation sets and automated checks so every change to prompts, retrieval or models is measured before it ships.
- Add guardrails for [sensitive data, unsafe output, actions the model may take] and decide where a person must review the output.
- Keep the features fast and affordable: caching, model choice, fallbacks between providers, cost per request.
- Monitor quality in production, investigate failures and turn them into new test cases.
- Work with product and design on what the feature should and shouldn't do.
What you'll need (must-haves)
- You've shipped at least one LLM feature that real users relied on, and can explain what changed after launch.
- You've measured quality systematically, with evaluation sets, metrics or structured human review, not just by reading outputs.
- You've built retrieval: chunking, embeddings, vector or hybrid search, re-ranking.
- Strong general software engineering in [Python / TypeScript]: tests, code review, production services.
- You've made and explained trade-offs between quality, latency and cost.
Nice to have
- Fine-tuning experience, for cases where prompting and retrieval aren't enough.
- Experience with [your domain, e.g. regulated data, multilingual users].
- Experience with [specific infrastructure you use].
How we work
[Two or three sentences on how decisions are made, how the team ships, and what on-call or support looks like.]
Pay and benefits
[Pay range, as required where you hire, and the main benefits.]
Interview process
- A 30-minute introductory conversation.
- A deep dive on an AI feature you built (60 minutes).
- A short work sample: [2 hours at most / 90 minutes live]. You can use AI coding tools.
- A system design conversation about an AI feature (60 minutes).
- A conversation with [a future peer / the product manager].
We aim to give you feedback within [number] working days of each stage.
Variants by company type
The template changes most in the "what you'll do" and "must-haves" sections:
| Company type | Emphasise | Add to must-haves | Watch out for |
|---|---|---|---|
| Early-stage startup | Breadth and speed: the first AI features, end to end | Full-stack or backend shipping experience; comfort choosing tools | Asking one person to also own model training and the data platform |
| Scale-up with an existing AI feature | Quality, reliability and cost at growing usage | Evaluation in CI; production monitoring; cost reduction | Hiding the hard part (an unreliable feature already in production) |
| Enterprise or regulated industry | Data handling, auditability, human review | Sensitive-data handling; working with security and legal | A long list of compliance buzzwords instead of the actual constraint |
| Agent-heavy product | Tool design, permissions and failure recovery | Built agents that take actions; designed limits on what the model may do | Over-promising autonomy in the ad itself |
Mistakes that attract the wrong people
- Title inflation or vagueness. "AI engineer" with an ML training job underneath attracts the wrong half of the market. Match the title to the work.
- "10 years of LLM experience." Modern LLM application work is only a few years old. Experienced candidates read this as a sign the team doesn't understand the field.
- Every framework at once. Listing every library in the ecosystem says you haven't decided what the person will build. Name your actual stack and let the rest go.
- No quality bar. If the ad never mentions evaluation, candidates who care about quality assume you don't.
- Hiding the constraints. If latency, cost or sensitive data are the hard part, say so. The right people are attracted by the real problem.
- Degree requirements by default. Unless the work truly needs it, ask for evidence of skills instead. See our guide to skills-based hiring for engineering teams.
Pay transparency and inclusive language
In many places, the pay range is now a legal requirement, not a choice. The EU Pay Transparency Directive gives job applicants the right to information about the starting pay or its range before the interview, bars employers from asking about pay history, and requires vacancy notices and job titles to be gender-neutral. Member states had to transpose it by 7 June 2026 [3]. Poland, for example, amended its Labour Code so that from 24 December 2025 employers must give candidates the proposed pay or range, may not ask about previous pay and must use gender-neutral job titles and ads [5]. In the US, several states require pay ranges in job postings, including California for employers with 15 or more employees [4]. Check the rules where the role is based.
Plain, specific language also widens the pool. A well-known study by Gaucher, Friesen and Kay found that job ads with more masculine-coded wording, such as "competitive" and "dominant", were rated as less appealing by women, and that women anticipated less belonging in those jobs [2]. The practical fix is to describe the work and the team concretely, and drop superlatives and combat or sports metaphors.
How Tailored Minds helps
When you brief a role on Tailored Minds, we turn a short description into a precise role profile in your company portal. This template is the public version of that thinking. We help you decide which AI role you actually need, using the LLM engineer profile and the other role definitions on our site. Then our team searches the talent network while approved independent recruiters can work the role at the same time, and people review every candidate we shortlist against the signals in the must-haves above. Hire AI engineers with Tailored Minds, or read our interview framework for AI engineers for the next step.
Frequently asked questions
What does an LLM engineer do?
What should an LLM engineer job description include?
How many years of experience should an LLM engineer have?
Should I title the role AI engineer or LLM engineer?
Do I have to include a salary range in the job description?
Sources
- LinkedIn Jobs on the Rise 2026: the fastest-growing jobs in the U.S. (reported by CNBC), CNBC
- Evidence that gendered wording in job advertisements exists and sustains gender inequality (Gaucher, Friesen and Kay, 2011), Journal of Personality and Social Psychology
- Directive (EU) 2023/970 (Pay Transparency Directive), Articles 5 and 34, EUR-Lex
- California Labor Code section 432.3, California Legislative Information
- Poland: Pay transparency measures employers must implement by 24 December 2025, Ogletree
About the author

Tercio Lima
Growth Lead at Tailored Minds
Tercio Lima is the Growth Lead at Tailored Minds. He runs the brand, content and search work that brings companies to the platform, and the sourcing work that finds engineers for their roles, so what he writes about hiring comes from live searches rather than desk research.
Share this article
