Building a Running Agent
Implementing intent recognition, task decomposition, planning, execution, tool calling, memory, and state management core links.
End-to-end responsible: problem definition, data and tool integration, prototype implementation, online and offline validation, evaluation, and reflection, output conclusion (continue or stop).
Designing and implementing Agent Runtime: session and task state machine, concurrency and queue, timeout and retry, permission and auditing, downgrade and rollback.
Collaborating with multiple roles: transforming requirements into executable agent spec, converting model capabilities into stable system capabilities.
Driving agent technology roadmap and key architecture decisions: tool protocol, memory scheme, evaluation metric, deployment and gray release mechanism.
Requirements
Bachelors degree or above in Computer Software Engineering or AI-related fields, with at least 3 years of experience in backend platform AI engineering or mature agent LLM application deployment experience.
Proficient in at least one major language (Python, Go, Java, etc.), with experience in large-scale project engineering.
Solid software engineering skills: architecture design, reliability, observability, performance and cost optimization, maintainability.
Complete delivery experience: able to turn exploration into a running system, with output conclusions based on evaluation data.
Good cross-team collaboration and expression skills: able to clearly describe trade-offs, boundaries, and risk factors.
Technical Experience
Agent Workflow Systems: • Multi-tool orchestration • Function call plugins • Task queue and scheduling • Long-term task and state recovery
RAG Knowledge Library Engineering: • Indexing • Retrieval • Re-ranking • Caching • Incremental update • Evaluation and observability
Multi-Modal Agent Experience: • Familiarity with multi-modal agents (image, speech, video) or experience with agent-based exploration in content search recommendation scenarios.
Public Influence: • Published papers • Patents • Open-source contributions • Industry influence (Agent LLM Systems and Infra)