About
The lab builds resettable, interactive digital environments where frontier AI agents learn by actively operating real software. It develops data infrastructure that enables models to train reliably on large, complex real-world datasets and collaborates with leading AI research labs.
Responsibilities
•Architect, build, and operate scalable distributed systems to acquire, ingest, and process large, unstructured datasets for agent training environments
•Design and implement automated data quality, validation, sanitisation, and PII-handling pipelines resilient to adversarial and messy real-world data
•Make foundational trade-offs across storage tiers, query performance, reliability, observability, and infrastructure cost without being tied to a fixed vendor stack
•Partner with platform, software, and AI research teams to establish platform-wide architectural patterns
•Own core data systems from initial problem definition through production operations
•Write production code, debug complex distributed systems across platform and AI engineering boundaries, and design validation, sanitisation, and ingestion infrastructure for agent training
Requirements
•Proven track record of owning substantial production data systems end to end, with personal accountability for architecture, implementation, and operational outcomes rather than just configuring existing ETL tools
•Strong, recent hands-on software engineering fundamentals and ability to work in unfamiliar codebases and debug distributed systems failure modes
•Demonstrated experience tackling large, complex, or messy datasets and articulating failure modes, trade-offs, and architectural alternatives
•Comfortable operating in high ambiguity: taking an unformed problem, deciding what to build, and executing without a detailed specification
•Based in or willing to work with engineering teams in London or Paris
Conditions
•Work location: London or Paris
