The University of Cambridge is recruiting a Research Assistant or Research Associate to work on trustworthy generative AI for computational science within the Department of Computer Science and Technology. The two-year fixed-term post focuses on a problem that is becoming increasingly important in scientific computing: large language models can generate code quickly, but incorrect or unverifiable code can undermine the reliability of scientific results.
The successful researcher will investigate ways to combine generative AI with software-engineering and programming-language techniques such as testing, static analysis, type systems, formal proof and domain-specific validation. The main scientific context will be climate and earth sciences. Salary is listed at £34,610–£35,608 for Research Assistant level and £37,694–£46,049 for Research Associate level. Applications close on 6 September 2026.
Position Summary
- University: University of Cambridge
- Department: Computer Science and Technology
- Reference: NR50505
- Location: West Cambridge, United Kingdom
- Contract: Fixed term
- Funding available for: 2 years
- Research Assistant salary: £34,610–£35,608
- Research Associate salary: £37,694–£46,049
- Published: 23 July 2026
- Closing date: 6 September 2026
The Research Problem Cambridge Wants to Solve
Programming is essential across modern science, from numerical modelling and simulation to data analysis and scientific software. Generative AI and large language models can potentially increase productivity by helping researchers write, translate, debug and reason about code.
The problem is trust. An LLM can produce code that looks plausible but is incorrect, numerically unstable, scientifically inappropriate or inconsistent with the assumptions of the underlying model. In climate and earth science, errors can affect large computational pipelines and the conclusions drawn from them.
Cambridge therefore wants to investigate how generative AI can support scientific software engineering without weakening scientific understanding or reliability. The project also takes a critical view of negative externalities, including environmental and human impacts, rather than treating AI adoption as automatically beneficial.
Potential Project Area 1: LLMs with Testing and Static Analysis
One possible research direction is to build frameworks in which LLM-generated code is not accepted at face value. Instead, analytical tools could sit inside a feedback loop and test or constrain the output.
Examples identified by Cambridge include conventional software testing, static analysis, type systems and climate-ensemble validators. A model might propose code, an analytical tool identifies a problem, and that feedback is then used to revise the generated output. This type of workflow could make AI-assisted scientific programming more reliable than simple prompt-and-copy use.
Potential Project Area 2: Generate Code Together with Proofs
Another direction involves formal verification. The project may investigate whether LLMs can produce code alongside machine-checkable proofs using systems such as Lean, Rocq or Agda.
The objective is not only to generate software but to provide stronger evidence that important properties of the computational model hold. For researchers with interests in programming languages, theorem proving, verification or formal methods, this is one of the most technically distinctive parts of the vacancy.
Potential Project Area 3: Translating Scientific Code
Large scientific communities still rely on older languages and codebases. Cambridge gives the example of Fortran, which remains widely used in climate science, and suggests research into AI-enabled translation toward modern environments such as Python or JAX.
The challenge is more complicated than translating syntax. Scientific code often contains assumptions about numerical precision, data layouts, parallelism, model behaviour and performance. A trustworthy translation system would need to preserve scientifically meaningful behaviour rather than merely produce code that runs.
Institute of Computing for Climate Science
The post sits within Cambridge’s Institute of Computing for Climate Science (ICCS), a multidisciplinary initiative that brings computer science, mathematics and software engineering into climate modelling. The Institute receives funding from several sources, including Schmidt Sciences.
The researcher will join a group applying programming-language and software-engineering research to scientific work. Cambridge also notes that there may be opportunities to collaborate with Research Software Engineers to turn research ideas into practical engineering artefacts.
Who Can Apply?
Candidates need degree-level education and should have a PhD, or be nearing completion of a PhD, in Computer Science, Engineering, Mathematics, Physical Sciences, Natural Sciences or a closely related field.
The vacancy requires research experience in at least one of the following:
- Computational modelling.
- Programming languages.
- Software or formal verification.
- Generative AI.
Applicants also need a track record of research and publications. Because this is a research appointment rather than an entry-level software role, evidence of independent or substantial academic research is important.
Desirable Experience
Cambridge lists three additional characteristics that can strengthen an application:
- Experience working directly with scientists from another domain.
- Experience turning research ideas into practical tools.
- Familiarity with generative-AI workflows and tools.
A candidate does not necessarily need all three. For example, someone with strong formal-methods research could be competitive even without deep climate-science experience if they can demonstrate an ability to work across disciplines.
Research Assistant vs Research Associate
The advert is open at both Research Assistant and Research Associate levels, with different salary bands. The final level normally depends on academic qualifications and experience. Applicants nearing completion of a PhD may be considered at the appropriate level, while a completed doctorate and relevant research record are more consistent with Research Associate expectations.
Applicants should not assume the top of the £46,049 band automatically applies. Appointment salary and grade are determined through the University’s recruitment process.
Application Documents
The Cambridge application requires more than a CV. Candidates should be ready to provide:
- A full curriculum vitae.
- A publications list.
- Contact details for two academic referees.
- A description of recent research, current research and future research interests for this role, limited to two pages.
Reference NR50505 should be quoted in the application and in correspondence about the vacancy.
How to Write the Two-Page Research Statement
The strongest research statement will connect the applicant’s existing work directly with the problem Cambridge is hiring for. Instead of spending most of the space describing a PhD thesis chronologically, identify the methods, results and skills that are relevant to trustworthy AI-assisted scientific software.
A programming-languages candidate could explain work on type systems, verification, program synthesis or theorem proving and then propose how those ideas could constrain LLM-generated scientific code. A generative-AI researcher could focus on code generation, evaluation, agents or tool use and show how they would introduce stronger reliability guarantees. A computational-science applicant could demonstrate knowledge of scientific modelling and explain what trustworthy AI assistance would need to preserve.
The future-research section should be concrete enough to show technical judgement but open enough to fit a collaborative research programme.
How to Present Your Publication Record
Do not rely only on journal prestige or citation counts. Make it easy for the panel to understand which publications show the skills needed for this post. Research on program analysis, formal methods, code generation, scientific machine learning, computational modelling, AI evaluation, scientific software or interdisciplinary collaboration should be particularly visible.
If a major paper is preprint-only or under review, label its status accurately.
Working Across Computer Science and Climate Science
This role is technically interdisciplinary. A researcher may need to understand the concerns of climate or earth scientists, collaborate with software engineers and still pursue publishable computer-science research.
Applicants who have previously worked with domain experts should explain what that collaboration required: translating requirements, validating models, maintaining research software, interpreting numerical output or reconciling different research cultures.
UK Work Eligibility
The University states that employees must be eligible to live and work in the United Kingdom. The public vacancy does not make a specific promise about visa sponsorship, so international applicants should check the University’s current immigration and right-to-work guidance alongside the vacancy before making assumptions about sponsorship.
How to Apply
Applications are submitted through the University of Cambridge recruitment system. Create or sign in to an applicant account, complete the required fields, upload the CV, publications list and two-page research statement, and provide details for two academic referees.
The deadline is 6 September 2026. Questions about the role or application process can be directed to the contact named in the official vacancy.
Frequently Asked Questions
How long is the Cambridge Generative AI research post?
Funding for the fixed-term appointment is initially available for two years.
What is the salary?
The listed Research Assistant band is £34,610–£35,608 and the Research Associate band is £37,694–£46,049.
Is a PhD required?
Applicants should have a PhD or be nearing completion in a relevant discipline such as Computer Science, Engineering, Mathematics or the physical or natural sciences.
Is this a general LLM research job?
No. The focus is specifically on trustworthy generative AI for scientific software engineering, primarily in climate and earth sciences.
