Agentic AI Research Engineer: Scale Intelligent Systems

AI Startups UK; Greater London; Hybrid; GBP 120,000 - 180,000; Full time

Benefits offered by this job

Competitive Salary

Equity & Ownership

Private healthcare

Visa sponsorship & relocation

London office

Job summary

Callosum in London seeks researchers to design and run experiments isolating why agentic AI systems fail, and to develop interventions that improve performance, cost and reliability across diverse silicon and models.

You'll publish findings, build instrumentation, and work across teams to ensure decisions are evidence-based rather than taste, with broad latitude to tackle problems at unprecedented scale.

Qualifications

• Proven ability to design open questions and run experiments that settle them.
• Experience with LLMs in agentic settings: multi-step tasks and tool use.
• Real experimental discipline: hypotheses, ablations, controlled comparisons.
• Strong engineering; able to contribute code within a shared repo.
• Clear written and verbal communication to translate results for teams.

Responsibilities

• Design and run experiments that isolate why agentic systems fail and propose interventions.
• Attack core research themes: memory management, agent steering, task decomposition.
• Build agent analysis and evaluation infrastructure for durable company use.
• Publish evaluation results and methods as company assets and public evidence.
• Collaborate across teams to design scalable systems at unprecedented scale and cost.

Skills

Research experience

Agentic LLMs

Experimental discipline

Engineering proficiency

Communication skills

Job description

Callosum in London seeks researchers to design and run experiments isolating why agentic AI systems fail, and to develop interventions that improve performance, cost and reliability across diverse silicon and models.

You'll publish findings, build instrumentation, and work across teams to ensure decisions are evidence-based rather than taste, with broad latitude to tackle problems at unprecedented scale.