Applied Scientist
i3 · London, England ·
Generative AI & Agentic AI | Insurance | London
London - hybrid, 3 days per week in the office
Previous insurance or reinsurance experience is required, ideally with an understanding of underwriting, claims, actuarial or other insurance data.
This is a senior, hands-on position for an experienced Applied Scientist who wants to take scientific ownership of sophisticated AI solutions, with a particular focus on LLMs, Generative AI and agentic systems.
You'll act as a senior technical authority within the Data Science & AI team, setting the scientific approach and quality standards for AI solutions from initial problem framing and experimentation through to evaluation, validation and production monitoring.
Importantly, this isn't a role where you'll move away from the technology. You'll remain hands-on, tackling complex problems, developing models and AI solutions, writing production-quality Python and reviewing the work of other scientists and engineers.
What you'll be doing:
- Leading the scientific design and development of AI/ML, Generative AI and agentic AI solutions
- Designing LLM applications incorporating RAG, tool use, prompting and multi-step agent workflows
- Owning evaluation methodologies for LLMs, RAG and agentic systems, including metrics, test sets and acceptance thresholds
- Applying strong mathematical and statistical rigour to experimentation, uncertainty, model behaviour and validation
- Solving complex problems involving unstructured documents, expert workflow automation, forecasting, optimisation, portfolio analysis and claims
- Setting standards for experimentation, evaluation, coding and documentation
- Leading Responsible AI approaches covering bias, fairness, explainability and model risk
- Writing production-quality Python and contributing directly to higher-risk or novel AI projects
- Working with Databricks, MLflow and Spark alongside modern LLM and AI frameworks
- Mentoring Data Scientists and Engineers through technical reviews, pairing and scientific guidance
- Acting as the escalation point for challenging scientific and modelling decisions
We're looking for:
- You'll need deep, hands-on experience with LLMs and Generative AI, including RAG, prompt engineering, fine-tuning and the design and evaluation of agentic AI systems.
- You'll also need excellent Python skills and strong foundations across statistics, probability, machine learning, experimental design and model validation.
- Crucially, you should have experience developing robust evaluation frameworks for AI systems - particularly measuring the quality and reliability of LLM, RAG and agent outputs - rather than simply building prototypes.
- Experience taking AI solutions from experimentation through to production is essential, alongside practical exposure to technologies such as Databricks, MLflow and Spark.
- Previous insurance or reinsurance experience is required, ideally with an understanding of underwriting, claims, actuarial or other insurance data.
- A postgraduate qualification in Computer Science, Statistics, Mathematics or another quantitative discipline would be advantageous.