Research Engineer, Agent Systems

We are looking for a Research Engineer to develop persistent agents for tasks that require memory, planning and recovery. You will build the software that lets an agent retain useful context, use tools and resume work when a task spans multiple sessions or encounters a failure.

These questions also matter for robots carrying out extended tasks: keeping track of progress, deciding what to do next and recovering from interruptions. You will advance the agent infrastructure and investigate how it can support task-level robotics workflows. The role combines stateful software engineering with empirical work on agent behaviour.

Responsibilities

• Develop durable task state, memory and execution mechanisms that preserve progress across interruptions and restarts.
• Build tool integrations with clear inputs, permissions and results, including appropriate handling of timeouts and partial failures.
• Design recovery and retry behaviour for concurrent work, with particular attention to preventing unintended repeated actions.
• Build evaluations for task completion, memory use and recovery, and create traces that make failures understandable and reproducible.
• Improve context management, execution latency and resource use without losing information needed for the task.
• Work with robotics researchers to define and test interfaces between task-level agents, simulation environments and robot capabilities.

Qualifications

• BS, MS or PhD in computer science, engineering or a related field, or demonstrated systems and research work of equivalent depth.

• Strong software engineering experience with stateful services, asynchronous execution, workflow systems or agent infrastructure.

• A solid understanding of concurrency, persistence and failure recovery, with examples of systems you have personally designed and debugged.

• Practical experience integrating models, external APIs or execution tools into a working application.

• The ability to analyse agent behaviour across prompts, model outputs, stored state and tool execution, and turn the analysis into a useful evaluation.

• Careful judgement about permissions and side effects, along with strong testing and technical communication skills.

Preferred experience

• Persistent agents, memory systems, workflow engines or tools that execute code and interact with user environments.
• Agent benchmarks, task verification, sandboxing or observability for complex software execution.
• Task planning, human supervision or interfaces between software agents and physical systems.
• Training or adapting agent models, reinforcement learning from interaction, or constructing tasks for behavioural evaluation.

Application materials

Include your CV and an agent or stateful system you have built. Explain the architecture, your contribution and a difficult failure you resolved. We are interested in how you evaluated reliability over repeated runs, as well as what the system could accomplish.