Research software engineers (RSEs) build the computational tools, data pipelines, and scientific software that make modern research possible: simulation frameworks for climate models, data analysis pipelines for genomics studies, high-performance computing infrastructure for physics experiments, and ML models that accelerate drug discovery. RSE salaries range from $118,000 at research universities to $188,000+ at national labs and private research organizations — with scientific computing engineers at industry research labs averaging $160,000.
This guide covers the RSE role, required skills, salary benchmarks, and how to navigate the research software engineering job market in 2026.
What does a research software engineer do?
Research software engineers write, maintain, and improve the software that scientists use to do research. This ranges from building the data processing pipelines that analyze telescope observations to developing the agent frameworks that AI research teams use to run experiments at scale. The defining characteristic of RSE work is that the software is itself a scientific instrument — its correctness, reproducibility, and performance directly determine the validity of research results.
Computational science engineers build simulation and modeling software
Computational scientists and engineers build the software frameworks that simulate physical, biological, or economic systems: finite element analysis tools for materials science, molecular dynamics simulations for drug discovery, climate models for atmospheric research, and agent-based models for epidemiology. This work requires deep mathematical understanding (numerical methods, linear algebra, differential equations) alongside the software engineering skills to make these simulations run correctly and efficiently on large computing clusters.
Data engineering for scientific pipelines is a distinct discipline
Scientific data pipelines differ from business data pipelines in important ways: data volumes can be enormous (petabytes from telescope arrays or genomics sequencers), data formats are domain-specific and often binary (HDF5, NetCDF, FITS files), analysis workflows are exploratory rather than production-defined, and reproducibility is a first-class requirement — the same pipeline must produce the same result on different hardware years later. Engineers who understand these constraints and can build workflows with scientific reproducibility requirements are specifically valuable in research contexts.
ML engineering for science (MLSci) is one of the fastest-growing RSE niches
Machine learning is accelerating science across domains: protein structure prediction (AlphaFold), materials discovery (GNoME), climate downscaling, drug candidate screening, and particle physics event classification. Engineers who can adapt ML methods to scientific problems — working with irregularly structured scientific data, handling physics-informed constraints in models, building uncertainty quantification into predictions — have a skill set that's increasingly in demand at national labs, pharmaceutical companies, and research universities.
What skills do research software engineering roles require?
Python is the universal language; C++ and Fortran for performance-critical code
Python is the primary language of modern research software: NumPy, SciPy, pandas, xarray, and the broader scientific Python ecosystem are the default tools for data analysis and model prototyping. For performance-critical code — numerical solvers, simulation kernels, high-throughput data processing — C++ and sometimes Fortran (a significant legacy codebase still powers critical weather and climate models) are used, often wrapped with Python bindings (pybind11, Cython). HPC-specific knowledge — MPI for distributed memory parallelism, OpenMP for shared memory, GPU programming with CUDA or ROCm — is required for the most computationally intensive roles.
Scientific software engineering practices: version control, testing, documentation
Research software has historically been written without the engineering discipline that production software requires. RSEs bring software engineering rigor to research: Git and version control, automated testing frameworks for numerical code (pytest with numerical tolerance fixtures), continuous integration, containerization (Docker, Singularity for HPC), and documentation standards that make software usable by other researchers. This combination — domain science context plus software engineering discipline — is the core RSE value proposition.
Domain specialization opens specific employer categories
Research software engineering is strongly domain-specific. Bioinformatics RSEs work with genomic data formats (FASTQ, BAM, VCF), sequence analysis workflows (Snakemake, Nextflow), and genomics databases (Ensembl, NCBI). Climate RSEs work with earth system model frameworks (CESM, GFDL), geospatial data libraries, and high-performance netCDF workflows. Physics RSEs might specialize in the CERN software stack (ROOT, Geant4) or gravitational wave analysis (LIGO). Domain knowledge is required in the most specialized research roles and is valued at generalist roles as contextual signal.
What do research software engineers earn in 2026?
Salary benchmarks
Based on Glassdoor and Comparably data for 2026:
- Research Software Engineer at universities: $80,000–$130,000 (varies significantly by institution and location)
- Research Software Engineer at national labs (Argonne, NREL, Oak Ridge, Lawrence Berkeley): $100,000–$160,000
- Scientific Computing Engineer at private research organizations: $146,000–$244,000 (avg $160,700)
- Scientific Computer Software Engineer: $160,000–$220,000 (avg $188,100)
- RSE at AI research labs (Google DeepMind, OpenAI, Anthropic, Microsoft Research): $160,000–$280,000+
University RSE positions are the most available entry-level pathway but pay significantly below national labs and industry research. For engineers targeting maximum compensation, industry research labs (pharmaceutical, tech, energy) and AI research organizations are the primary target.
Industry perspective
"According to the 2024 RSE Survey conducted by the UK Software Sustainability Institute and the US Research Software Engineer Association, 83% of research software engineers hold a postgraduate degree — but the survey also found that 31% transitioned into RSE roles from software industry positions without research backgrounds, and that this group reported the highest salary satisfaction. The survey notes that RSE salaries at national laboratories and private research organizations have increased 18% over the prior two years, closing a gap that previously pushed engineers toward industry."
— Software Sustainability Institute RSE Survey 2024
How do you find research software engineering jobs?
Know the employer types and their tradeoffs
RSE roles exist across four employer categories with different compensation and culture:
- Universities: most common entry point, lowest pay, high intellectual freedom, access to research collaborations
- National laboratories (Argonne, NREL, Sandia, Lawrence Berkeley, CERN, Jülich): substantially better pay than universities, stability, world-class computing infrastructure
- Pharmaceutical and biotech R&D (Pfizer, AstraZeneca, Genentech, BioNTech): industry pay, domain focus on drug discovery and genomics
- AI research labs (Google DeepMind, OpenAI, Microsoft Research, Meta FAIR): highest pay, ML focus, intense competition for positions
For engineers entering from industry, national labs and pharmaceutical R&D offer the best combination of research environment and competitive compensation.
Research software engineer communities and job boards
The Research Software Engineers Association (UK) and US-RSE maintain job boards and community resources specifically for the RSE community. Academic job boards (HigherEdJobs, jobs.ac.uk for UK positions, AcademicJobsOnline) list university RSE positions. National laboratory jobs are posted through USAjobs.gov (for US national labs) or directly on lab career sites. For remote-eligible RSE roles filtered by research domain and timezone, browse Hire.monster's science and research industry feed.
Open source contribution and publication signal research software depth
For RSE roles, particularly at universities and national labs, evidence of research software work that has been used by others carries significant weight: contributions to major scientific Python projects (NumPy, SciPy, pandas, Astropy, scikit-learn), publications in the Journal of Open Source Software (JOSS), or documented software packages used in published research. This is the RSE equivalent of an industry portfolio — it demonstrates that your software was good enough to be adopted and cited. For tailoring a resume for RSE positions, listing software citations and GitHub repository usage statistics (stars, downstream dependents) alongside technical skills is more effective than generic software engineering bullets.
Key takeaways
The industry-to-RSE transition is increasingly valued and well-compensated
The RSE community has historically been composed of scientists who learned programming. The emerging pattern — software engineers who develop scientific domain knowledge — is now equally valued and consistently better compensated. National labs and industry research organizations actively recruit engineers from industry who can bring software engineering discipline to research contexts. For engineers with HPC, data engineering, or ML backgrounds, the transition to RSE roles at national labs or pharmaceutical R&D is a path to interesting problems with good compensation.
HPC and GPU programming are the RSE premium skills in 2026
As ML training and scientific simulation move to GPU accelerators, engineers who can write efficient CUDA kernels, optimize MPI-based parallel codes, or profile and tune code for specific HPC architectures are specifically scarce. The combination of high-performance computing knowledge and ML framework experience (PyTorch, JAX) is what national labs and AI research organizations most actively recruit for. This skill set is difficult to acquire without access to actual HPC hardware, making research organization alumni particularly valuable.
Scientific reproducibility requirements are the unique engineering constraint in RSE
Production software must be correct and reliable. Research software must be correct, reliable, and reproducible — years later, on different hardware, by different researchers. This means specific engineering practices: containerizing computational environments, storing random seeds, using deterministic algorithms where possible, archiving input data with DOIs, and writing documentation that explains not just how the software works but why specific algorithmic choices were made. Engineers who understand this reproducibility requirement (and can implement Snakemake or Nextflow workflows that enforce it) are prepared for research environments; engineers who don't are surprised by it.
Frequently asked questions
Do I need a PhD to work as a research software engineer?
It depends on the employer. University RSE positions often prefer or require a graduate degree, particularly for positions embedded in specific research groups. National lab and industry RSE positions are more varied — many explicitly hire BSc and MSc holders with strong software engineering experience. AI research lab positions at the research engineer level generally don't require a PhD. The trend across all employer types is toward weighting demonstrated software engineering skills more heavily than academic credentials.
What is HPC and do I need it for RSE roles?
HPC (High-Performance Computing) refers to computing environments that use large clusters of processors — often thousands of CPUs and GPUs — to run computations that would take too long on a single machine. Many scientific simulations (climate models, molecular dynamics, particle physics) require HPC. HPC knowledge (SLURM job scheduler, MPI parallelism, efficient parallel I/O) is required for RSE roles embedded in compute-intensive research groups but is not needed for all RSE positions — bioinformatics RSEs, software engineering roles in computational social science, or RSEs building data visualization tools may not interact with HPC clusters.
What's the difference between a research scientist and a research software engineer?
A research scientist formulates and tests scientific hypotheses — their output is knowledge and publications. A research software engineer builds the tools that scientists use to do this work — their output is software. In practice the roles overlap at the margins: RSEs who work closely with researchers develop scientific insight; some research scientists write substantial code. The clearest distinction is that RSE roles are evaluated on software quality metrics (code that works, scales, and is maintainable) while research scientist roles are evaluated on scientific contribution (publications, patents, discoveries).
Are remote RSE roles available?
Yes, increasingly. National labs have expanded remote and hybrid arrangements since 2020. Pharmaceutical R&D RSE roles have substantial remote availability. AI research lab RSE positions at major tech companies are often fully remote. University positions are more likely to require on-campus presence, particularly for RSEs embedded in specific lab groups. Remote RSE availability is expanding faster than it is in traditional academic research roles, driven by the engineering (not scientific) nature of the work.
What are good ways to build research software engineering skills before applying?
Contributing to open-source scientific Python projects (NumPy, SciPy, pandas, domain-specific packages like Astropy or BioPython) is the highest-signal portfolio activity. Building a Snakemake or Nextflow workflow for a publicly available dataset demonstrates research workflow engineering. Taking free HPC courses (LLNL, XSEDE training materials) and documenting the projects gives concrete resume content. JOSS (Journal of Open Source Software) publishes peer-reviewed software papers — submitting a research software tool through JOSS produces a citable publication that signals RSE-specific accomplishment.
Bottom line
- RSE salaries range from $80K at universities to $188K+ at scientific computing and AI research organizations
- Python for analysis + C++/CUDA for performance + HPC knowledge is the premium RSE skill combination
- National labs, pharmaceutical R&D, and AI research labs pay far more than universities for equivalent skills
- Scientific reproducibility (containerized workflows, deterministic pipelines) is the unique engineering constraint in this domain
- Browse research software engineering roles on Hire.monster