Machine Learning Engineer

The post holder is a member of the R&D team, reporting to the company’s Technical Director, and leads the development, training, and ongoing improvement of the AI agent that operates our automated experimental platform. The role combines hands‑on machine learning engineering with ownership of the training strategy — defining what data the agent needs, how that data is collected and structured, how the agent is trained and evaluated, and how it is adapted to improve real‑world performance.

The post holder is highly skilled in machine learning engineering, with strong knowledge of agent architectures, training pipelines, and evaluation methodology, and is comfortable working with models that control physical hardware rather than operating on static datasets alone. The person will work closely with our neuroscience, hardware engineers, and software team to translate domain expertise into training objectives, data specifications, and measurable performance criteria.

• Design, build, and maintain the AI agent that plans and executes automated experimental workflows on the platform.
• Select and implement appropriate learning approaches (e.g. imitation learning, reinforcement learning, supervised policy learning, or hybrid rule-and-model architectures) for each stage of the workflow.
• Define the agent’s action space, observation space, decision rules, and success and failure conditions in collaboration with domain experts.
• Develop and maintain training, fine‑tuning, and deployment pipelines, including model versioning and reproducibility.
• Integrate the agent with real‑time control software, sensors, and instrument hardware in partnership with the software and engineering teams.
• Build simulation or offline test environments where these reduce reliance on scarce or costly experimental runs.

• Specify what data is required to train the agent effectively — including data types, sampling rates, annotation schemes, volume, and coverage of edge cases and failure modes.
• Work with experimental staff to design and implement data collection protocols that produce clean, well‑structured, and reusable training datasets.
• Build and maintain data pipelines for ingestion, labelling, quality control, versioning, and storage.
• Identify gaps, imbalances, and biases in the training data, and recommend targeted collection to address them.
• Continuously assess the return on additional data versus changes to model architecture, training regime, or reward and objective design.

• Define and maintain evaluation metrics and benchmark test suites that reflect real experimental success, not proxy measures alone.
• Analyse agent performance, diagnose failure modes, and identify their root causes in the model, the data, or the surrounding system.
• Adjust and retrain the agent to improve success rate, consistency, and robustness across different samples, operators, and hardware configurations.
• Monitor deployed agent performance and establish processes for detecting drift and triggering retraining.
• Document model behaviour, limitations, and known failure conditions to a standard suitable for a regulated or quality‑controlled product environment.

• Contribute to strategic decisions on AI and software architecture for the current developing system and planned future capabilities.
• Manage and monitor machine learning resources, compute, and budgets in relevant areas.
• Support deployment, troubleshooting, or performance investigation of the agent at customer or partner sites as required.
• Contribute to technical documentation, patents, and reports as needed.
• Train or mentor other team members on machine learning methods, tooling, or data practices where relevant.