Software Engineer: AI Harnesses & Tooling

Role overview Build the software layer that makes language models dependable, extensible agents. The work spans runtime behavior, memory, collaboration, retrieval, model selection, and permissions, with ownership of problems that help agents preserve context and act safely over long running tasks.

Responsibilities - Build the agent runtime, including task dispatch, tool calls, execution state, retries, cancellation, and error handling. - Design an extension system that lets users, field engineers, and agents adapt the harness to practical needs. - Define how agents store learned information with provenance, confidence, and a way to handle changing or stale facts. - Support coordination among agents through concurrent work, partial failures, retries, and restarts. - Develop local retrieval for personal documents and messages, keeping context current and responsive on a user’s device. - Implement model routing and trust controls, including provider fallback, sandboxing, permissions, and auditable actions.

Requirements - The source does not provide a formal qualifications list; it describes a hands on software engineering role building agent infrastructure. - Ability to design reliable systems that handle retries, failures, and long running execution. - Interest or experience in language model agents, tool use, retrieval, or extensible runtimes. - Sound judgment around permissions, credentials, user approval, and auditability. - Ability to take ownership of technical problems across multiple areas of the product.

Benefits and work setup - Full time engineering role, remote within US or EU time zones. - Early team position with stated US base salary of $215,000–$325,000 and equity.