LLM Engineer Jobs

LLM engineer jobs on carbon.talent, also listed as generative AI or GenAI engineer roles. These positions focus specifically on large language models: retrieval-augmented generation, fine-tuning, evaluation, agents and inference optimisation. They overlap with AI engineering but tend to go deeper into how models behave and how to adapt them. This guide covers the skills that recur in listings, how interviews probe them, and how to present work that is more than a demo.

Open roles

LLM engineering splits into application work and model work. Application-focused roles build retrieval pipelines, tool-using agents and structured-output systems on top of hosted or open models. Model-focused roles fine-tune, distil or quantise models, run inference servers and optimise throughput. Some listings want both. The phrases 'fine-tuning', 'inference' and 'GPU' signal model work; 'RAG', 'agents' and 'product' signal application work.

Retrieval is a core skill for application roles. Expect questions on chunking strategies, embedding choice, hybrid search, re-ranking, and how to evaluate retrieval separately from generation. A common failure in interviews is treating RAG as one step; strong candidates measure recall of the retriever and faithfulness of the answer separately and know which to fix first.

For model-focused roles, expect PyTorch, Hugging Face tooling, parameter-efficient fine-tuning methods such as LoRA, dataset curation, and serving frameworks designed for LLM inference. Understanding memory, batching and quantisation trade-offs helps you answer questions about cost and latency. Data quality for fine-tuning, including deduplication and contamination checks against evaluation sets, comes up often.

Evaluation is central in both tracks. Be ready to discuss building task-specific test sets, using models to grade outputs and when that is unreliable, human review protocols, and tracking regressions when a model or prompt changes. Interviewers want to hear how you decide an output is good enough to ship.

A credible project shows measurement. Fine-tune a small open model on a narrow task and compare it with a prompted baseline on a held-out set, or build a retrieval system with separate retrieval and answer metrics and a short write-up of what improved them. Publish the evaluation code. Projects without numbers are hard for hiring managers to assess.

LLM engineering is offered on full-time, contract and fractional terms. Contract work often covers a specific build, such as an internal assistant over company documents, while fractional roles help companies decide which problems suit LLMs at all. Remote roles are widespread, though model-focused roles may require access to specific compute environments.

Safety and data handling come up in most interviews: preventing prompt injection in agents, keeping personal data out of logs, and controlling which tools a model can call. Employers in finance, health and legal sectors will ask about audit trails.

Questions to ask an employer: which models are in use and why; whether you will work with hosted APIs, open models or both; how evaluation sets are built and refreshed; what the latency and cost budgets are; and who signs off on releases. Teams without answers are often early in their adoption, which can mean more influence and more ambiguity.

Applications are weakest when they list frameworks and models without measured results. Write about one system in depth: the task, the baseline, what you changed, the metric that moved and the failure cases that remained. Mention cost and latency where you can. That level of detail is rare and memorable.

Stay current without chasing every release. Pick a small set of sources, re-run your own evaluation when a new model appears, and note what changed. Being able to say how a model update affected your task is far more useful in interview than knowing the latest benchmark score.

If you are moving from general software engineering, start with retrieval and evaluation, where engineering discipline transfers directly. Model-level work such as fine-tuning can follow once you can measure quality reliably.

Open-source contributions to evaluation tools or inference libraries are a strong signal, because they are reviewed in public by people who know the field.

Questions to ask an employer: which models are in use and why; whether you will work with hosted APIs, open models or both; how evaluation sets are built and refreshed; what the latency and cost budgets are; and who signs off on releases. Teams without answers are often early in their adoption, which can mean more influence and more ambiguity.

Applications are weakest when they list frameworks and models without measured results. Write about one system in depth: the task, the baseline, what you changed, the metric that moved and the failure cases that remained. Mention cost and latency where you can. That level of detail is rare and memorable.

Stay current without chasing every release. Pick a small set of sources, re-run your own evaluation when a new model appears, and note what changed. Being able to say how a model update affected your task is far more useful in interview than knowing the latest benchmark score.

If you are moving from general software engineering, start with retrieval and evaluation, where engineering discipline transfers directly. Model-level work such as fine-tuning can follow once you can measure quality reliably.

Open-source contributions to evaluation tools or inference libraries are a strong signal, because they are reviewed in public by people who know the field.

carbon.talent lists LLM engineer roles with employment type and locations. A free profile lets you show the systems you have built and the work types you want.

How carbon.talent works for candidates

  • Free signup, profile and applying.
  • Contract, fractional and full-time data and AI roles.
  • carbon.talent takes no cut of a contractor's rate.
Create your free profile

Questions

What is the difference between an LLM engineer and an AI engineer?

They overlap heavily. LLM engineer roles focus specifically on large language models and often go deeper into retrieval, fine-tuning, evaluation and inference; AI engineer is a broader title.

What skills do LLM engineer jobs require?

Python, retrieval and search, evaluation methods, and for model-focused roles PyTorch, fine-tuning techniques and inference optimisation.

How do I show LLM experience without a job in it?

Build a small project with a held-out evaluation set and publish the results and code, for example a fine-tuned model compared with a prompted baseline.

Are LLM engineer jobs remote?

Many are. Check each listing's hiring countries; model-focused roles may have compute or security requirements.