Agent Harness 工程师
Core
Design and implement the core framework for Agent runtime, including scheduling, tool abstraction, context management, and observability systems to provide a stable execution base for business applications.
Role type
Senior IC machine-learning engineer (AI Agent infrastructure)
Builds
Scalable Agent execution platforms, observability pipelines, automated evaluation systems, and debugging toolchains for LLM-based agents.
Domain
Artificial Intelligence / Large Language Models / Agent Systems
Deliverable
production ML models
Required skills
Python, TypeScript or Go, system design, distributed systems, high-concurrency services, asynchronous programming, LLM inference and sampling, Agent frameworks (LangGraph/AutoGen/MCP), Tool Calling, Function Schema design, Prompt and Context Engineering, RAG, memory mechanisms, OpenTelemetry, trace storage and query, metric system design, offline data processing pipelines.
Preferred skills
Experience with agentic coding tools (Cursor/Claude Code/Codex), deep understanding of Agent failure modes, open source contributions, complex system refactoring, vertical domain Agent deployment (Code/Kernel/Performance/Data Analysis), RL/SFT data loop construction.
Responsibilities
Architect and implement the Agent runtime framework with scheduling and sandboxing; build full-link tracing and observability for LLM and tool calls; construct automated evaluation pipelines with benchmarks and A/B testing; develop debugging toolchains for trace replay and failure reproduction; participate in building core support platforms for data and model iteration.
Seniority
Senior, hands-on IC
