Summary

Research Engineer focused on designing rigorous evaluations for agentic systems, including multi-turn, tool-using, long-horizon llm behaviour. This is an in-person role at Callosum's London office, building evaluation methodology, observability infrastructure, and evidence-based methods for improving ai systems.

Responsibilities

• Design benchmarks for agentic behaviour
• Build statistical evaluation methodology and harnesses
• Red-team evaluations and design sanity checks
• Turn raw traces into failure taxonomies
• Attribute outcomes to causal decisions
• Infer capabilities from partial evidence
• Build evaluation and observability infrastructure