Research Engineer (AI Safety)

TL;DR

Research Engineer (AI Safety): Building and maintaining an internal suite of benchmarks for content and agentic guardrails with an accent on model capabilities and agent behaviors. Focus on quantifying realistic LLM failure modes in the wild and creating defensible capability benchmarks.

Location: Hybrid in Paris or London. Relocation package available for Paris.

Compensation: $150K – $250K + Equity

Company

An AI Safety company building the safety, reliability, and optimization layer for AI systems through natural-language policies.

What you will do

• Own and maintain the internal benchmark suite for single/multi-turn content guardrails and agentic safety.
• Develop benchmarks that distinguish specific model capabilities and create evals for flagship models.
• Build benchmarks for new features emerging from the research team.
• Adapt and extend evaluations to new verticals and evolving product data.
• Execute research projects to study and quantify realistic agentic and LLM failure modes in the wild.

Requirements

• Experience building LLM benchmarks from scratch that produced measurable and defensible capability differences.
• Proven track record of building synthetic data for post-training textual or multimodal models.
• Ability to reproduce published benchmark results and identify methodological fragilities.
• Strong Python skills with experience shipping and maintaining production-grade code.
• Expertise in efficient LLM inference setups, including orchestration, parallel calls, and rate-limit handling.
• Fluency with frontier models and coding agents on a daily basis.

Nice to have

• Experience with automated red-teaming.
• Experience working across various agentic scaffolds and reproducing public benchmarks.
• Deep knowledge of existing reward-model, monitoring, or safety benchmarks.
• Published papers in the evaluations or safety-evaluation domain.

Culture & Benefits

• Paid time off according to local regulations.
• Relocation support for Paris-based roles.
• Comprehensive medical insurance for the France-based team.
• Provision of all necessary hardware, tools, and covered subscriptions for AI agents and IDEs.
• Bi-annual team off-sites in various locations.

Hiring process

• Introductory call with HR (25 min).
• Take-home technical test task.
• Technical interview with the Head of Fundamental Research (60 min).
• Final conversation with the CEO (45 min).