Leadenhall Analytics
Website:
leadenhallanalytics.com
Company:
https://www.linkedin.com/company/leadenhallanalytics
Seniority: Entry level
Industries: Insurance
Job details:
About Us
Leadenhall Analytics is a specialist consultancy providing actuarial and exposure management expertise to the London Market, working with Lloyd’s syndicates, (re)insurers, and MGAs. Founded by experienced London Market actuaries, the firm focuses on solving complex analytical problems and implementing practical, production-ready solutions.
We are a small team, and deliberately so. Everyone here does work that reaches a client.
We are building AI-powered applications: internal tooling that makes our own consultants faster, and products aimed at the market we serve.
Role Description
You will work directly with the Founders to turn ideas into working applications. Not prototypes that sit in a folder, things that get used.
Because we are small, the work is genuinely varied. In a given month you might be wiring up an LLM pipeline over a pile of unstructured insurance documents, building the interface someone uses to interrogate it, working out why the retrieval quality is poor, and then sitting in on a call to hear what the user actually wanted.
This suits someone who wants breadth and ownership early. It does not suit someone who wants a narrow, well-defined ticket queue.
What you will be doing
- Building applications end to end, from data and backend logic through to the front end
- Working with frontier LLMs: prompting, evaluation, retrieval, tool use, agent patterns
- Fine-tuning smaller open-weight models for our specific tasks
- Building and curating the training and evaluation datasets that work depends on
- Designing and iterating on interfaces that non-technical insurance professionals will actually use
- Testing your own work properly, including building evaluation harnesses for AI features where deterministic tests do not apply
- Deploying and maintaining what you build
- Talking to users and translating vague requirements into something buildable
What we are looking for
- 0 to 2 years of commercial experience, plus a degree in computer science, engineering, mathematics or a comparably technical subject
- Strong Python
- Working knowledge of TypeScript and React, or clear evidence you can pick them up fast
- Hands-on experience building with LLMs, whether professionally, at university, or in your own projects. Personal projects count and we will look at them
- Fluent with agentic coding tools such as Claude Code or Codex. We expect you to work with them, not around them, and to have a view on where they help and where they do not
- Practical exposure to fine-tuning open-weight models. LoRA or QLoRA, a training framework such as Axolotl, Unsloth or TRL, and enough understanding of evaluation to know whether your fine-tune actually improved anything
- Comfortable with Git, and with deploying something to a real environment
- Able to work independently across a time zone gap, and to tell us early when you are stuck rather than late
- Clear written English. Most of our communication is asynchronous and written
Useful but not required
- Postgres, Supabase, or similar
- Cloud deployment experience
- Any exposure to insurance, actuarial work or financial services. We will teach you the domain, and we would rather hire for engineering ability and curiosity
What you get
- Direct access to the people making decisions. No layers
- Real ownership of what you build, with your name on it
- Domain training in a specialist, well-paid field that very few engineers understand
- Access to dedicated hardware for training and inference work, so you are not fighting for cloud credits every time you want to run something
- Remote working
We want to ensure that all applicants have a fair and equal chance, so we’re doing an initial assessment to minimise unconscious bias in our hiring process. Successful candidates will be invited to a job interview.
Ready to join our team? Start by clicking the "Apply" link above.
Click on Apply to know more.
