AI Engineer — open roles, skills, and who's hiring
There are 260 open AI/ML engineering roles across 42 companies tracked by the Agentic Ready Jobs Index, as of 12 June 2026. That is 30% of the 866 agentic roles in the index, second only to agent engineering, and 43 of the 260 (17%) are remote — roughly double the remote share of the agent engineering category. Anthropic alone accounts for 64 of the openings.
What an AI engineer is
An AI engineer builds applications and systems on top of machine learning models — most often, in 2026, large language models. The role covers the ground between research and product: taking a model that works in a notebook and making it work in a system, with retrieval, fine-tuning, evaluation, and serving decisions along the way. Where an agent engineer specialises in the decide-act-observe loop, an AI engineer's scope is broader and closer to the model itself.
The title hides at least three different jobs, and the current postings show all three. Applied AI engineers (48 of the 260 postings carry "Applied AI" in the title) integrate existing models into products and customer systems. Research engineers (111 postings carry "Research" in the title — the largest slice) work on training, post-training, and evaluation inside labs, where "engineer" means building the infrastructure that makes research run. And classic ML engineers build and maintain the platforms — feature pipelines, training systems, model serving — that everything else depends on. Read the team name in the posting, not the title, to know which one you are applying for.
This is the category where labs dominate. Anthropic, Scale AI, Mistral AI, and OpenAI hold the top four spots, and lab postings lean research-side: post-training, reinforcement learning, voice and speech, data infrastructure. The non-lab postings — Datadog, Reddit, Instacart — lean applied-side: AI features inside an existing product.
Skills and tools that appear in real postings
Seniority skews higher here than in agent engineering: 52 of 260 postings (20%) are staff-plus, against 9% in the agent category.
How to break in
The applied end is the realistic entry point: it values production engineering plus demonstrated model integration, and 143 of the 260 postings (55%) are mid-level. The research-engineer end mostly hires from infrastructure and systems backgrounds or from research itself, and it concentrates at the labs. If your background is general backend engineering, an applied AI role at a product company is the shorter jump; the lab research roles tend to follow a few years later, if at all.
Adjacent roles: agent engineer for the agent-loop specialisation, agent ops engineer for the serving and platform side, and forward deployed engineer if you want customer-facing applied work. The full dataset is on the Agentic AI Jobs Index.
Skills appearing in real postings
Hiring for this role right now
- Anthropic 57 roles San Francisco Careers ↗
- Netflix 38 roles Los Gatos CA Careers ↗
- Adobe 35 roles San Jose Careers ↗
- Capital One 34 roles McLean VA Careers ↗
- General Motors 33 roles Detroit Careers ↗
Live from the Agentic AI Jobs Index, updated 16 June 2026.
Salary
Levels.fyi lists a median of $151,000 for the "AI Engineer" title and $243,000 for its broader ML/AI software engineer category — the gap reflects how loosely the title is applied, with the lower figure including many rebadged developer roles. Both figures are self-reported and skew toward large US employers. None of the 260 postings in the 12 June 2026 snapshot disclosed a range, so the index cannot yet publish posting-based numbers for this category.
Sources: levels.fyi — AI engineer · levels.fyi — ML/AI software engineer