AI Engineer Job Description
An AI engineer job description template: the ATS keywords recruiters screen for, a realistic salary range, and how the role differs from Machine Learning Engineer.
"AI Engineer" is a title that barely existed before large language models went mainstream in late 2022, and demand for it grew faster than any shared definition of what the job actually is. At most companies today, the work is applied software engineering against a model someone else trained — prompting, retrieval, evaluation, orchestration — not model architecture or gradient descent. If you're picturing a research role, this usually isn't it.
The title collides constantly with Machine Learning Engineer, and two postings can describe nearly the same work or completely different work depending on the company. Where a product is built on top of third-party foundation models, "AI Engineer" functions as a specialized backend role — the differentiators are retrieval design, prompt and evaluation infrastructure, and the cost and latency tradeoffs of API calls, not model internals. Where a company trains or fine-tunes its own models, the AI Engineer title is sometimes just Machine Learning Engineer with a newer-sounding name on the same job. Read the requirements bullets, not the title, to know which one you're looking at.
Worth knowing before you apply: this field's best practices are still moving every few months, so a listed requirement like a specific framework often means "comfortable learning whatever the current tool is" more than deep expertise in that exact one. A company that hired for a particular framework in 2023 may have rebuilt most of that code since. Candidates coming from traditional software engineering with a few shipped LLM-powered features tend to screen just as well as candidates with a formal ML background — this is one of the few senior technical titles in 2026 where the on-ramp is still genuinely open.
Bellamyre · Raleigh, NC
Full-time · Hybrid
$155,000 – $205,000
About the role
Bellamyre is adding a natural-language layer on top of its forecasting platform — the ability for a customer to ask a plain-English question about their own data and get a grounded, correct answer back, not a plausible-sounding guess. We're hiring a senior AI Engineer to build that layer: retrieval, prompting, evaluation, and the guardrails that keep it honest.
This role sits closer to product engineering than to model research. You won't be training a foundation model from scratch; you'll be building the retrieval pipelines, prompt and evaluation infrastructure, and agent workflows that make a third-party model useful and trustworthy against our specific data. Based in Raleigh, hybrid with the rest of the AI Platform team.
What you'll do
- Design and build product features powered by large language models — natural-language search, generated summaries, conversational interfaces — from prototype through production.
- Build and maintain retrieval-augmented generation pipelines: chunking, embedding, indexing and retrieval against our own data.
- Write, version and evaluate prompts systematically, against test sets and metrics rather than by eyeballing outputs.
- Integrate third-party model APIs — and self-hosted open-weight models where it makes sense — while managing cost, latency and rate limits across providers.
- Build tool-calling and agentic workflows for tasks that need a model to take multi-step actions, with guardrails against unsafe or runaway behavior.
- Monitor deployed AI features for output quality and hallucination rate, and iterate on prompts or retrieval when quality drifts.
- Partner with product and design on what a model should — and should not — be trusted to do unsupervised.
- Track a fast-moving vendor landscape, since a single new model release can change what's worth building versus buying within a quarter.
What we're looking for
- Four or more years of software engineering experience, including at least a year or two shipping features built on large language models.
- Strong Python (or a comparable backend language) and experience calling and orchestrating third-party APIs.
- Practical experience with retrieval-augmented generation: embeddings, vector search, chunking strategy.
- A track record of building evaluation harnesses for non-deterministic model output, not just shipping and hoping.
- Working knowledge of prompt engineering as a discipline — structured prompting, few-shot examples, output parsing and validation.
- Comfort reasoning about the cost and latency tradeoffs of model calls at production scale.
Nice to have
- Experience fine-tuning or self-hosting an open-weight model, such as Llama or Mistral.
- Familiarity with vector databases — Pinecone, Weaviate, pgvector — at production scale.
- Background in traditional machine learning or data science, on teams that blend both disciplines.
- Experience building multi-agent or tool-using AI workflows.
- Exposure to model safety and guardrail patterns — content filtering, output validation, human-in-the-loop review.
Benefits
- Medical, dental and vision coverage, with employee premiums covered in full.
- 401(k) with a company match, vested immediately.
- Hybrid schedule: two days a week in the Raleigh office with the AI Platform team.
- Dedicated budget for model API usage, tooling and conference attendance.
- Twenty days of paid time off plus company holidays.
Salary range
As posted for this sample role. Real pay varies by employer, location and experience.
$155,000–$205,000/ yr
ATS keywords for this role
The applicant tracking system (ATS) — the recruiting software a hiring team searches and filters applicants with — will screen for these. Weight shows how central each one is to this specific posting.
Required and central (4)
Important (6)
Mentioned in passing (9)
Frequently asked questions
What's the difference between an AI Engineer and a Machine Learning Engineer?
Machine Learning Engineer more often means training and serving your own models. AI Engineer more often means building product features on top of a foundation model someone else trained — the work is closer to backend engineering than to model research. The titles get used inconsistently enough that the requirements section is a better guide than either one.
Do I need to know how to train a neural network to get an AI Engineer job?
Usually not. Most AI Engineer roles are API-first: the daily skill is retrieval design, prompt engineering and evaluation, not gradient descent. If a posting titled AI Engineer spends most of its requirements on training infrastructure, it's effectively a Machine Learning Engineer role with a newer title.
Is this a stable field to build a career in, or will the tools change under me?
The tools will keep changing — frameworks and best practices in this space have shifted every few months since 2023, and that isn't likely to stop soon. The transferable part is the underlying judgment: how to evaluate non-deterministic output, design retrieval that actually grounds a model, and reason about failure modes. That skill set carries across whatever framework happens to be current.
This was a sample. Your resume should be tailored to the real thing.
Rezi Ninja reads an actual job posting and rewrites your resume to match it, with a Ninja Score so you know it lands before you hit send. Free to start, with the AI usage included.