Senior ML Research Engineer — Agentic Systems & AI Infra

Callosum; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

Callosum, in London, seeks researchers to design experiments that diagnose why agentic AI systems fail and to turn insights into interventions. You will work on memory management, agent steering, and continual learning, building durable analysis tools used across the company.

The role offers hands-on work with multi-agent systems, strong emphasis on experimental discipline, and opportunities to publish results both internally and publicly. This is a London-based, on-site role.

Qualifications

• Evidence that you can run research of your own: design experiments and produce reusable results.
• Hands-on experience with LLMs in agentic settings: multi-step tasks, tool use, long horizons.
• Real experimental discipline with hypotheses, comparisons, ablations and uncertainty.
• Strong engineering to implement and run experiments within a shared codebase.
• Strong communication to translate research into guidance for teams and wider audiences.

Responsibilities

• Design and run experiments isolating why agentic systems fail and turn findings into interventions.
• Attack core research themes: context/memory management, agent steering, task decomposition and specialization; continual learning; inter-agent communication.
• Build agent analysis and evaluation infrastructure: observability, behaviour analysis, evaluation harnesses, auditing.
• Publish evaluation results and discoveries as company assets and public evidence.
• Collaborate across teams to design scaleable, low-cost, low-latency systems; explore novel models and silicon.

Skills

Research design

LLMs in agents

Experimental discipline

Strong engineering

Communication skills

Job description

Callosum, in London, seeks researchers to design experiments that diagnose why agentic AI systems fail and to turn insights into interventions. You will work on memory management, agent steering, and continual learning, building durable analysis tools used across the company.

The role offers hands-on work with multi-agent systems, strong emphasis on experimental discipline, and opportunities to publish results both internally and publicly. This is a London-based, on-site role.