Job description
Project overview
Data Scientist
Inside IR35: £600 - £650 p/d
Active SC required
Initially till March 2027, rolling extensions from there
Responsibilities
What you'll bring
Bonus points: Inspect (AISI's evaluation framework), promptfoo, LLM-as-a-judge, or security data such as scanner findings, CVEs and SIEM alerts.
Good to know: this role is about applying models, not training or fine-tuning them.
- Design and run head-to-head tests of frontier models (Claude, OpenAI, Gemini and sovereign options) on real cyber use cases
- Build the test datasets and scoring that show which model wins for each job, and why
- Measure what matters: accuracy, false positives, cost and speed
- Turn your results into clear evidence senior leaders can act on
- Work alongside AI engineers scaling the pilot into a live product
- Active SC clearance (essential)
- Hands-on experience evaluating LLMs; you've tested them, not just used them
- Strong Python and statistics, and you're at home with messy real-world data
- You take ownership and thrive with light-touch oversight
- A genuine interest in cyber security
Project details
occasional travel to London
Which AI model is best at finding a vulnerability before an attacker does? Come and prove it.
You'll join a central government cyber security team at the heart of the UK's new push to use frontier AI for cyber defence. Their AI code-scanning pilot is already finding real weaknesses across government. Now they need you to turn it into hard evidence.
- Start: ASAP
- Duration: 6 Months +
- Location: London, UK, GB
- Rate: 75-81 EUR/h
Indexed by 9am's meta-search. You are a job provider? Claim or manage this listing
Found on JobServe