Remote AI Jobs — 12 Specialist Roles, One Curated Guide
The AI job market has fractured into a dozen specialisations in the last three years. "AI engineer" and "ML engineer" mean different things at different companies. Research and applied science diverge as you go senior. A new track — AI product management and AI safety research — has emerged alongside them.
This page maps the current landscape, with a dedicated guide for each role covering what the job actually involves, what employers expect, and what separates strong candidates. Every guide links straight through to live remote listings.
The three AI role tracks
Most AI roles sort into one of three tracks. Engineering ships the systems, research pushes the frontier, and the product+prompt track sits between model capability and the user. Picking the track first makes the role decision much easier.
Ship the systems that serve AI in production. Highest headcount, broadest remote hiring, most diverse employer pool.
AI Engineer
→Application-layer ML. Building features on top of frontier models with evaluation and guardrails.
ML Engineer
→Production ML systems — training pipelines, serving infra, observability, monitoring.
LLM Engineer
→LLM specialist track. Prompting, fine-tuning, inference optimisation, and LLM-specific tooling.
RAG Engineer
→Retrieval-augmented systems. Indexing, chunking, retrieval quality, grounded generation.
Agent Engineer
→Tool-using LLM agents — planning loops, action-space design, evaluation of autonomous behaviour.
MLOps Engineer
→Infrastructure and pipelines for ML at scale. Feature stores, model registries, CI/CD for models.
Push ML capability, adapt models to product, and study the behaviour and safety of frontier systems. PhD-heavy; frontier-lab-heavy.
Applied Scientist
→The bridge between research and product. Model adaptation, evaluation design, fine-tuning.
Research Scientist
→Original ML research. Experiments, papers, new methods. Frontier labs and industrial research.
AI Safety Researcher
→Alignment, interpretability, red-teaming, and evaluations of frontier AI capabilities.
Data Scientist
→Business-analytical ML. Experimentation, causal inference, decision-support modelling.
Sit between model capability and user experience. The newest and fastest-growing specialisation inside product management.
AI salary snapshot — US total compensation
Typical US total compensation bands across the six most common AI role tracks. Frontier-lab compensation at senior+ levels routinely exceeds these ranges. European numbers are typically 20–35% below; frontier-lab EU roles often close the gap.
| Role | Mid (3–6 yrs) | Senior (6–10 yrs) | Staff / Principal |
|---|---|---|---|
| AI Engineer | $170K–$240K | $230K–$340K | $310K–$480K |
| ML Engineer | $180K–$260K | $240K–$360K | $320K–$500K |
| Applied Scientist | $200K–$280K | $300K–$440K | $430K–$680K |
| Research Scientist | $240K–$340K | $360K–$520K | $550K–$1.0M+ |
| AI Safety Researcher | $250K–$360K | $380K–$560K | $560K–$1.1M |
| AI Product Manager | $180K–$270K | $240K–$360K | $320K–$480K |
Bands are drawn from the individual role guides — see each for methodology and level definitions.
Where AI roles live — four employer types
The same role looks different at each of these employer types. Understanding the employer matters more than memorising job descriptions.
Frontier AI Labs
Anthropic · OpenAI · Google DeepMind · Meta FAIR · Mistral · Cohere
Top-of-market compensation; research and safety depth; frontier-adjacent engineering.
AI-Native Application Companies
Perplexity · Harvey · Glean · Hebbia · Replit · Cursor · Character.AI
Entire product is AI-powered. Engineers own model behaviour end-to-end.
Big Tech AI Product Orgs
Google · Microsoft · Meta · Amazon · Apple
Large formal teams. Deep specialisation. Clear career ladder.
SaaS Companies Adding AI
Notion · Figma · Linear · Slack · HubSpot · Intercom · Atlassian
AI features inside existing product. Most common 2026 hiring pattern.
Which AI role is right for you?
If you have strong software engineering fundamentals and want to ship AI features fast →
If you own production ML infrastructure — training, serving, observability →
If you have a PhD or equivalent research output and want depth over breadth →
If you shape product and want to work closer to AI model behaviour →
Frequently asked questions
What is the difference between an AI engineer and an ML engineer?
AI engineers typically work at the application layer, building features on top of frontier models from OpenAI, Anthropic, or open-weight providers — heavy on integration, evaluation, and UX. ML engineers typically own the underlying system — training pipelines, model serving, infrastructure, observability. The line is drawn differently at every company, so read the listing carefully.
Which AI role pays the most?
At senior levels, research scientist and AI safety researcher roles at frontier labs (Anthropic, OpenAI, Google DeepMind) regularly pay the most, often $500K–$1M+ total compensation. Applied scientist and ML engineer roles at frontier labs are close behind. At non-lab companies, AI engineer and ML engineer roles in the $250K–$450K band are more typical for senior individual contributors.
Do I need a PhD to work in AI?
Not for most roles. Engineering-track roles (AI engineer, LLM engineer, RAG engineer, agent engineer, MLOps engineer, AI product manager, prompt engineer) very rarely require a PhD. Applied scientist roles at frontier labs typically do or accept equivalent research output. Research scientist and AI safety researcher roles almost always require a PhD or demonstrable top-venue publication history.
Which AI track has the best remote hiring?
Engineering-track AI roles are the most remote-friendly across the board — the work is laptop-and-cluster, document-heavy, and async-compatible. Research and applied-science roles have meaningful remote hiring too, though some frontier-lab teams still prefer hybrid for specific capability or safety workstreams. Product and prompt roles are fully remote-compatible.
How do I pivot from software engineering into AI?
The most efficient path is through AI engineer or LLM engineer roles — the gap is narrower because you keep your engineering fundamentals and add ML literacy, evaluation design, and prompt/model familiarity. Three to six months of serious side-project work (building a real AI product, reading papers, shipping public evaluations) usually closes the gap for engineering-track roles. For applied-scientist or research roles the bar is higher.
Which companies are hiring remote AI talent in 2026?
Frontier labs (Anthropic, OpenAI, Google DeepMind, Meta, Mistral, Cohere) hire meaningfully remote for most roles. AI-native application companies (Perplexity, Harvey, Glean, Replit, Cursor, Character.AI) are hiring heavily. Traditional SaaS companies adding AI features (Notion, Figma, Linear, Slack, HubSpot, Intercom) hire AI engineers and AI PMs. Big tech AI product orgs (Google, Microsoft, Meta, Amazon) continue to expand.
Ready to apply?
RemNavi aggregates remote AI jobs from Jobicy, Remote OK, We Work Remotely, Remotive, and direct employer pages. Every listing links straight through to the employer.