AI research job titles, decoded

What does each title actually mean?

The same person can be a research scientist in Zurich, an applied scientist in Seattle and a member of technical staff in San Francisco without their work changing. The table reads the titles as employers use them, not as a career ladder.

How the common titles are used, and what to check before believing one.

Title	Where you see it	What it usually means	What to verify
Research scientist	Frontier labs, industrial research groups, some universities	Owns an open question. Publication expected, or at least allowed	Whether publication is genuinely permitted, and who signs off
AI scientist	Advertisements more than business cards	The same role, written for a non technical reader	Nothing extra. Treat it as a synonym for research scientist
Applied scientist	Amazon and AWS, and companies that copied the ladder	Research work tied to a production system	The ratio of experiment to deployment, which varies wildly
Member of technical staff	OpenAI, Anthropic, and labs that copied them	Everyone technical, from first hire to staff level	Everything. The title carries no seniority and no specialism
Research engineer	Frontier labs, and universities with a different meaning	Builds what the experiments run on. At a lab, often senior	Whether the person shaped the research question or served it
Research fellow	Mostly British and Commonwealth universities	A postdoc on a fixed term contract	When the contract ends, which is when they become reachable
Machine learning engineer	Product companies of every size	Ships models. Rarely proposes methods	Whether you need a method at all, or an implementation

Is an AI scientist the same as an AI researcher?

In hiring, yes. The two phrases describe one job, and the person doing it is almost always titled research scientist on the organisation chart. Companies write AI scientist in advertisements because it reads plainly to a hiring manager who is not from the field.

The phrase has a second life outside hiring, and it interferes. Since 2024 the AI scientist also names software that generates hypotheses, runs experiments and writes up results without a person in the loop. Search for the term and you get papers about that, which is worth knowing before you put it in an advertisement: candidates searching for work do not use it, and your posting lands among tool reviews.

How do you read seniority when the title hides it?

Member of technical staff removes the ladder on purpose, and it works: two people holding it can be five years apart. The record does not hide the same way. Count papers where your candidate is first author at venues with a hard review process, and read the co-author list to see whether they worked inside a large group or led something.

The date of a first publication sets the floor on experience, and it is the field we index most completely. A first paper in 2019 and a defence dated 2024 describes a doctoral trajectory; a first paper in 2012 with a gap in the middle usually describes someone who moved into industry and came back. How to find and hire AI researchers sets out the ranking in detail, and the guide to NeurIPS, ICML and ICLR authors says what an acceptance at each venue certifies.

Where do the people holding these titles work?

Company titles only apply to people who are in a company, and most indexed researchers are not. Among the profiles whose current position we have read, 18% name an employer that resolves to a company. The rest sit in universities and public laboratories, where the titles are doctoral student, postdoctoral researcher, lecturer and professor, and where the moment that matters is the end of a contract rather than a promotion.

The move from one world to the other is dated, and it is a better signal than any title. The table below counts the year each indexed researcher started at a company, from what they declared themselves.

Which title should you advertise?

Write research scientist when the role owns an open question and publication is allowed. Write research engineer when the person will build the training and evaluation machinery, and say so plainly rather than dressing it as research. Write machine learning engineer when the method exists and the work is to make it run.

The title decides who reads past the first line, and the first line decides who answers. Researchers read the problem before the ladder, so name the problem: the capability you are missing, the compute you will give, and whether the work can be published. An advertisement that oversells the research content is found out in week one, and the person leaves within the year.

If you are unsure which of the three jobs you need, the profile guide asks the one question that separates them.

Common questions

Is an AI scientist the same as an AI researcher?

In a company, yes. AI scientist, research scientist and AI researcher describe the same work: proposing a method, running the experiments that test it, and publishing or shipping the result. The title people actually hold is research scientist. Outside hiring, the phrase now also names software that runs experiments on its own, which is why a search for AI scientist returns papers about agents rather than people.

What is the difference between a research scientist and an applied scientist?

The employer, more than the work. Applied scientist is the Amazon title, used across AWS and the retail side, and it covers people who would be called research scientists elsewhere. The distinction that matters is whether the role owns an open question or a known method applied to a production system, and the title rarely tells you which.

What does member of technical staff mean at an AI lab?

One title for the whole technical population. OpenAI and Anthropic use it for researchers and engineers alike, which removes the internal hierarchy and makes an external read impossible: two people holding it can be three years apart in seniority and work on different halves of the stack. Judge the record rather than the title.

Which title should we advertise for an AI research role?

Research scientist, if the role owns an open question and publication is allowed. Research engineer, if the person will build the training and evaluation infrastructure. Naming the problem in the first line of the ad matters more than either: researchers read the problem before the title, and a title that oversells is discovered in week one.

Does a research engineer rank below a research scientist?

Not at a frontier lab. The people who make training runs work are scarce and paid accordingly, and several of the most cited results of the last three years carry a research engineer as first author. In a university the ladder is real and a research engineer sits under the academic staff, which is where the assumption comes from.

How do you tell a senior AI researcher from a junior one without the title?

Count first-authored papers at venues with a hard review process, and read who they published with. A doctoral student can hold four first-authored NeurIPS papers, and a director of research can have none from the last three years. Position in the author list and the date of the first publication say more than any title on a profile.