The People Between the Experiment and the AI: RIKEN’s New Research Engineer Fellowship

A scientific dataset can be available to everyone and still be difficult to use. A program can produce a promising result while leaving its assumptions obscure. Between obtaining a measurement and trusting a conclusion lies a substantial engineering task: making the methods, software and data intelligible enough to test again.

RIKEN is creating a fellowship for people able to work in that space. Recruitment for its Research Engineer Fellow programme opened on September 2, with an announcement published on September 4. The institution wants researchers with mathematical, computational and IT expertise to deepen their use of data and AI through practical scientific work. Its Information R&D and Strategy Headquarters will support skills development, including security. RIKEN presents wider careers in universities and industry as an eventual goal. ^[1]

The first intake, for fiscal 2027, is expected to comprise about five people. This is paid, fixed-term employment, rather than a student scholarship. Its significance lies less in the initial headcount than in the professional role it seeks to develop. ^[2]

A research partner with an engineering skill set

RIKEN’s programme description treats a research engineer as a participant in solving scientific problems, with the expertise to connect a research question to computational methods. Fellows are, in principle, to work across multiple laboratories during their appointment. That gives the programme a different emphasis from a placement devoted exclusively to one laboratory’s immediate needs. ^[4]

Consider a general example. An analysis appears to distinguish two groups of experimental samples. A technically sound program may have found a real biological difference—or a difference in how the samples were measured. Someone who understands both the experiment and its computational treatment can help decide what to check next. Conversely, knowing the scientific question is insufficient if the underlying files and processing steps cannot be reconstructed.

This is why the combination of expertise matters. It also explains why the fellowship should not be read as a promise to turn every participant into an expert in every discipline. The host list identifies different assignments and skills; it explicitly does not require every listed skill in every candidate. Matching a person’s experience to an agreed research task is part of the application process. ^[5]

What the work could involve

The host laboratory list ranges from research data infrastructure to life science and quantum computing. One team proposes work on collecting, managing, sharing and analysing research data, including automation and electronic laboratory notebooks. A biomedical data team includes preparation, standardisation and quality of data for AI and machine learning. A quantum computer systems team lists software work related to fault-tolerant quantum computation. These are examples of possible assignments, not a curriculum that every fellow will complete. ^[5]

The range matters because “using AI” describes only part of the task. Experimental measurements come with conditions and limitations that affect interpretation. Computational systems have assumptions that determine what a model represents. An engineer working with researchers must understand enough of the underlying science to recognise when a technically successful operation has failed to answer the question.

The history behind a new job description

Founded in 1917, RIKEN now spans fields including physics, chemistry, life science and computational science. On March 9, 2021, the Fugaku supercomputer began shared use. That combination of disciplinary breadth and computing infrastructure provides context for the fellowship, although the programme does not itself promise fellows an automatic allocation on Fugaku. ^[6][7]

There is also an international history of trying to establish careers around the software that research depends upon. The Society of Research Software Engineering traces its origins to discussions at a 2012 workshop in Britain about the role and careers of people developing research software. The society was established in 2019. ^[8]

RIKEN’s research engineer role and the Research Software Engineer profession are related, but their scope is not identical: the Japanese fellowship also encompasses data management and connections to experimental work. Japan.co.jp’s reading is that both address a shared institutional question—how to recognise technical expertise as a scientific contribution and sustain the careers of its practitioners. This comparison does not establish that RIKEN has adopted the British model.

From finding data to trusting its reuse

A second historical strand concerns the care of scientific data. The FAIR principles, formally published in 2016, set out the goals of making research objects findable, accessible, interoperable and reusable. They emphasise machine use as well as human use. Putting files online is consequently only one part of making them useful to future research. ^[9]

RIKEN offered a practical example of the discovery problem in February 2026, when it launched its Data Portal. The portal brings together information about roughly 70 publicly available databases and supports searches across research fields. It did not announce 70 newly created datasets or merge all the underlying data into one repository. It provided a common entrance to resources whose information had been dispersed. ^[10]

Discovery and reliable reuse are connected tasks, but they are different tasks. Finding a file does not establish whether its units, measurement conditions or processing history permit comparison with another file. Nor does it settle the conditions under which the information may be used. Maintaining that context becomes part of the foundation on which AI-assisted conclusions can be examined.

The application: two deadlines, a laboratory agreement

Applicants must discuss a placement and its work with a prospective host before registering. Documents and recommendation letters may be submitted in Japanese or English; two referees are required. Registration and the completed application have different deadlines. ^[2][3]

This is a summary. Applicants should consult the linked guidelines for the full skills requirements, exceptions, submission procedures and renewal provisions. ^[2][3]

Japan.co.jp analysis: what should survive the appointment?

Moving between laboratories could help fellows carry useful methods across disciplinary boundaries. It could also make continuity harder unless there is time to document work and transfer knowledge. The value of mobility will depend partly on whether a laboratory can keep using what a fellow has helped build.

That suggests a way to assess the programme beyond counting appointments. Reusable software, well-described data and procedures that another researcher can reproduce deserve attention alongside papers. This is an editorial proposal for judging outcomes, not a description of RIKEN’s published assessment criteria.

The career question remains equally consequential. RIKEN says a subsequent career path within the institution is under consideration; it has not guaranteed permanent employment after the fellowship. With the first intake still ahead, there are no completed fellowships from which to judge the programme’s results. ^[4]

The experiment is therefore institutional as well as technical. Can a limited appointment leave both a durable contribution to research and a stronger professional future for the person who made it? The answer will emerge in the systems that remain useful after a fellow moves on—and in the work that person is able to do next.