AI Software Engineer, Agent Harness
Core
Design and build the end-to-end agent harness architecture that runs agents against files, tools, documents, and memory with safe execution and permissions.
Role type
Senior IC AI Software Engineer (Agent Infrastructure)
Builds
Production agent harness components including session semantics, orchestrators, memory systems, and tool-use interfaces for open-weight models.
Domain
AI Infrastructure / Systems Engineering / Agent Orchestration
Deliverable
production ML models
Required skills
Python, Go/Rust/C++, agent loop design, RAG, context management, sandboxing, permission models, open-weight model serving, quantization, tool validation and repair, evaluation harness design, API/CLI/GUI development
Preferred skills
Open-source agent harness contributions, agent benchmarking (SWE-bench, Terminal-Bench), MoE serving, LLM observability, document parsing pipelines, desktop/GUI application experience
Technologies
Python, Go, Rust, C++, Open-weight models, MCP servers
Responsibilities
Own the harness architecture end-to-end; build unique components like session semantics and memory systems; keep pace with model changes via adapters and prompt formats; ensure reliable tool use across models; develop and review agents/tools/MCP servers; build evaluation harnesses; define interfaces (API, CLI, GUI, endpoints)
Seniority
Senior, hands-on IC