Staff Engineer: Context & Memory for AI Agents

scaleai; Greater London; On-site; GBP 120,000 - 180,000; Full time

Job summary

Scale AI's Applied Intelligence Systems (AIS) team in London is seeking a Staff Software Engineer to own hard problems end to end, from backend context services to the evaluation decisions that prove usefulness.

You will design core memory and retrieval primitives, work across distributed systems and applied ML, and set technical standards adopted by the team in collaboration with SWE and MLE peers.

Qualifications

• 8+ years of engineering experience building production systems end to end.
• Direct experience with ML and information retrieval problems: embeddings, vector search, retrieval augmented generation, fine-tuning, or memory architectures.
• Comfort owning both software engineering and applied ML breadth, with strong judgment to make calls.
• Proficiency in Python and production ML infra: containers, cloud platforms, CI/CD.
• Track record of setting technical standards and influencing decisions beyond the team.

Responsibilities

• Own large, ambiguous problems in context and memory end to end, from design through production.
• Architect primitives for retrieval, storage, and cross-session memory with knowledge bases and vector stores.
• Design and maintain evaluation rubrics and benchmarks for memory and retrieval quality.
• Set technical standards and patterns adopted across the team.
• Collaborate with SWE and MLE peers and other AIS workstreams on memory-context intersections.
• Debug production issues related to context, retrieval, and cross-tenant memory at scale.

Job description

Scale AI's Applied Intelligence Systems (AIS) team in London is seeking a Staff Software Engineer to own hard problems end to end, from backend context services to the evaluation decisions that prove usefulness.

You will design core memory and retrieval primitives, work across distributed systems and applied ML, and set technical standards adopted by the team in collaboration with SWE and MLE peers.