AI Research Engineer – Model Behaviour

Cager is redefining how organisations deploy AI agents. Our zero-ops platform runs each agent in an isolated environment, fully separated from core business systems, while anomaly detection identifies model misbehaviour before it causes harm. Available as SaaS or on-premises, Cager enables organisations to adopt AI agents safely and with confidence.

We aim for 10x, not 10%. Every role contributes to a future where prevention is the norm and the internet is safer for everyone.

Lead the research and development of Cager's anomaly detection. You will study how AI agents behave across models and turn complex, real-world activity into detection that performs reliably in production.

• Model-agnostic anomaly detection deployed across client environments.
• A rigorous evaluation framework covering false positives, missed incidents and time to detection.
• Detection tuned in close collaboration with pilot clients on live workloads.

• Extensive experience in applied machine learning, from research to production.
• A proven track record of making algorithms work under real-world conditions.
• Strong machine learning fundamentals and algorithmic thinking across multiple problem domains, with sound evaluation practice.
• Strong software engineering skills in Python, including production code, testing and code review.

• Experience in cross-functional R&D teams alongside platform and software engineers.
• Clear communication, with the ability to explain trade-offs to clients, engineers and management alike.
• Fluency with modern AI tooling to increase productivity and quality.

• Client first. Every decision starts with the needs of the people we serve.
• 10x, not 10%. We pursue step changes, not incremental gains.
• Technology-enabled, human-centred. We build advanced systems for real people.
• Ownership. We deliver production-quality work, using AI tools to move faster while keeping judgement and accountability our own.
• Shared knowledge. We mentor, review each other's work and foster a culture of openness and continuous improvement.

Compensation will be confirmed following our initial funding round.