AI Applied Scientist

This is a UK-based role that makes the AI systems which power our products measurably better. How well those systems perform is at the heart of what Aykua offers. You will design the evaluations that tell us how accurate they are, find out why they fail, and run the experiments that fix it. You will statistically assess prompt and model performance and maintain our internal evaluation frameworks. The role suits a scientist with rigorous experimental habits, from a PhD, postdoc or equivalent industry experience, who wants their work to ship in a product which real clinics are using today.

What You’ll Do

• Improve how we measure AI quality: extend the evaluation frameworks for our existing products and stand up new ones as products are built, assessing prompt and model performance statistically and designing experiments that compare changes rigorously.
• Build and continuously improve our LLM-as-judge evaluations: calibrate judges against expert human review, find where they fail, and feed what you learn back into our prompts, models and products.
• Curate and maintain our internal evaluation datasets, working with expert reviewers to establish and keep ground truth.
• Work closely with our product and engineering teams to turn your findings into improvements that reach production.
• Mentor the team in evaluation and experimental practice, establish best practices, and raise the technical bar for how we measure AI quality across the company.

• A PhD and postdoctoral research experience, or equivalent industry experience (5+ years), in data science, machine learning or a related quantitative field.
• Bachelor’s or Master’s degree in Computer Science, Data Science, Statistics, Mathematics, or a related quantitative field.
• Strong proficiency in programming languages such as Python (with libraries like Pydantic, Pandas, NumPy, Scikit-learn).
• A strong foundation in statistical analysis, experimental design and hypothesis testing, and experience applying it to large or messy datasets.
• Experience turning analysis into working tools: evaluation harnesses, data pipelines or reusable experiment code that others on the team can run.

You will be a great fit if you…

• Enjoy working with frontier models.
• Would rather measure whether something works than assume it does.
• Automate away repetitive analysis and build tooling that makes evaluation faster and more reliable.

It would be nice if you had…

• Hands-on experience with LLMs, such as prompting, LLM-as-judge or fine-tuning.
• Familiarity with evaluation metrics such as precision and recall, and with using them to check automated judgements against human review.
• Experience with healthcare or clinical data, or in another regulated industry, including de-identification, privacy-preserving data work or data governance for protected data (PII/PHI).
• Publications in peer-reviewed venues.
• Experience using coding agents in research work, with independent verification of the results.

Please note: Applications will be assessed on a rolling basis, and we aim to respond to every applicant within one week. You must have the right to work in the UK, as this is an in-person role near Cambridge.