AI Safety Benchmark Engineer

Moonfire; Greater London; Hybride; GBP 90 000 - 130 000; Plein temps

Avantages offerts par ce poste

Equity

Hybrid work model

Office in London/Paris

Relocation support

Private health insurance

Mental health support

Résumé du poste

White Circle in London is seeking a research engineer to build and maintain an internal benchmark suite spanning single/multi-turn content and agentic guardrails. You will collaborate with the core safety team to study agent behaviours and push evaluation tooling into production.

Applicants should have hands-on Python production experience, strong benchmark design skills, and a track record of shipping reliable code for ML systems. Hybrid work with offices in London and Paris is offered.

Qualifications

• Have built an LLM benchmark from scratch that distinguished specific model capabilities.
• Have built synthetic data for post-training textual or multimodal models.
• Can reproduce a published benchmark result and identify fragile methodologies.
• You write Python that other people can build on and ship to production.
• You can write efficient LLM inference setups with orchestration of parallel calls and retries.
• An AI power-user fluent with frontier models and coding agents day to day.

Responsabilités

• Own and maintain our internal benchmark suite for single/multi-turn content guardrails.
• Build benchmarks that distinguish model capabilities and feed into product evals.
• Adapt evals to new features and changing product data.
• Collaborate with product and research teams on evaluation tooling.
• Work on research projects studying agent behaviours in the wild.

Connaissances

Python

LLM benchmarking

Experiment design

Production-grade code

Distributed systems

Formation

Master's degree in CS/AI

Outils

Docker

Git

PyTorch

Description du poste

White Circle in London is seeking a research engineer to build and maintain an internal benchmark suite spanning single/multi-turn content and agentic guardrails. You will collaborate with the core safety team to study agent behaviours and push evaluation tooling into production.

Applicants should have hands-on Python production experience, strong benchmark design skills, and a track record of shipping reliable code for ML systems. Hybrid work with offices in London and Paris is offered.