2 months ago
London, United KingdomSenior
Responsibilities
- Design and run experiments across agent behavior, traces, and activations to identify failures and develop interventions that improve intelligence, cost, and reliability.
- Research context and memory management, agent steering, task decomposition and specialization, continual learning during deployment, and inter-agent communication.
- Build durable agent analysis and evaluation infrastructure, including observability, behavior analysis, evaluation harnesses, and auditing tools.
- Publish evaluation results, methods, and discoveries as company assets and public research.
- Collaborate across teams to design scalable, cost-efficient, low-latency systems and novel model or agent architectures for heterogeneous hardware.
Requirements
- Demonstrated ability to independently conduct research by formulating open questions, designing experiments, and producing reusable results.
- Deep hands-on experience with LLMs in agentic settings, including multi-step tasks, tool use, and long-horizon workflows.
- Strong experimental discipline with hypotheses, controlled comparisons, ablations, and careful treatment of uncertainty and benchmark noise.
- Strong software engineering skills and comfort writing experiment code within a substantial shared codebase.
- Strong communication skills for translating research findings into guidance and communicating them to wider audiences.
Benefits
- Competitive salary determined by skills and experience, equity and ownership, and private healthcare.
- Visa sponsorship and relocation benefits are offered.
- The role is based in person at the company's London office with provided tools, workspace, and setup.
Categories
AI Research
About Callosum
Callosum builds AI infrastructure that unifies heterogeneous compute—software and systems to orchestrate diverse accelerators and the physical clusters they run on—for organizations training and serving models. The privately held company was founded in 2025 and is headquartered in London. It focuses on designing, deploying, and operating mixed-hardware compute fleets across sites and vendors, and raised a seed round in 2026.