By Carla Burger, founder of PeopleNotResumes. Updated October 2026.
| ML engineer | Applied AI engineer | |
|---|---|---|
| Core job | Train, evaluate and deploy models on your data | Build features and agents on foundation models |
| Typical work | Feature pipelines, training runs, model serving, monitoring for drift | Prompting, retrieval, tool use, evaluation harnesses, cost and latency control |
| Starting point | Your data | An existing model and your product |
| Key skills | Statistics, ML frameworks, data engineering, MLOps | Software engineering, LLM APIs, retrieval, evaluation design |
| How you know it works | Offline metrics, then online tests against a baseline | Task-level evals, human review, production feedback |
| Common background | ML, data science or data engineering | Backend or full-stack engineering, then LLM products |
An ML engineer builds systems that learn from your data. That means preparing training data, choosing and training models, measuring them against a baseline, and running them reliably in production. Recommendation, ranking, fraud detection, forecasting and pricing are classic ML engineering problems. The work sits close to data engineering and infrastructure, and a good ML engineer cares as much about monitoring and retraining as about the first model. See how we recruit machine learning engineers.
An applied AI engineer, often called an LLM engineer or simply an AI engineer, builds products on top of foundation models that someone else trained. The work is retrieval, prompting, tool use and agents, wrapped in software that real users depend on. The hard parts are evaluation, reliability, cost and latency: knowing whether the system gives good answers, and keeping it fast and affordable as usage grows. Most applied AI engineers come from strong software engineering backgrounds. See how we recruit applied AI and LLM engineers.
Start from your product, not the title. If your product sends requests to a model API and your edge is in the workflow, the interface or the data you retrieve, hire an applied AI engineer. If your edge is a model trained on data nobody else has, hire an ML engineer. If you are an early-stage startup and need one person to build the whole product, you probably need a founding engineer with applied AI experience. Non-technical founders will find more help in our guide to hiring your first AI engineer.
Many teams end up with both. A common path: applied AI engineers ship the first version on a foundation model, then ML engineers join once there is enough usage data to fine-tune, train smaller models, or build ranking and personalisation. The two roles work best when they share one evaluation standard, so everyone agrees what "better" means.
For both roles, ask candidates to walk through something they shipped and how they knew it worked. For ML engineers, probe data quality, baselines, offline versus online metrics, and what happens when a model drifts. For applied AI engineers, probe evaluation design, retrieval quality, failure handling, and the trade-offs between cost, latency and accuracy. Avoid scoring candidates on lists of tools, which change every few months. Our AI engineer interview questions that test judgment cover both roles.
Name the problem the person will own, the users they will serve, and the stage of the system: a new build, a first version in production, or a mature system that needs to scale. Say which models, data and infrastructure they will work with. Be clear about whether you expect them to train models or build on existing ones, because that one line decides which candidates apply. Then use the title your target candidates use for themselves. For the full process, read our guide on how to hire AI engineers.
Not usually. In most job postings today, "AI engineer" means someone who builds products on foundation models, which is applied AI engineering. "ML engineer" usually means someone who trains and deploys models on a company's own data. Titles vary, so read the work, not the label.
Often, yes, especially if they have strong software engineering skills. The reverse is harder: an applied AI engineer without ML training may struggle with model training, data pipelines and statistical evaluation.
No. Many strong applied AI engineers come from software engineering. What matters is how they evaluate model output, handle failures and ship reliable systems.
Yes. We run ML and applied AI searches in San Francisco, New York, Austin and London, with behavioural assessment and a human approving every candidate decision.
Tell us what you are building and where the pipeline is stuck. You'll hear back within one business day with a scope, a timeline, and a fee.
Start a search