About the job
Patsnap is developing an advanced Agent Brain platform for AI, designed to support complex knowledge work in fields such as Intellectual Property and Research & Development. Unlike typical chat-based LLM interfaces, this project aims to build a specialized agent architecture capable of managing long-chain tasks, integrating varied data sources, and producing results that are both verifiable and evidence-based.
Role overview
The Lead / Staff Engineer, AI Agent Platform, will focus on creating a reusable infrastructure for agents and a robust reasoning and orchestration kernel. The role addresses three core technical challenges:
- Execution Engine: Develop the core mechanics behind agent loops, including multi-step reasoning, middleware pipelines, planning and orchestration of sub-agents, as well as features like checkpointing, state recovery, and execution controls such as permissions, cost tracking, and clarification handling.
- Context Engineering: Advance beyond simple context window management by optimizing input standardization, defining interaction capabilities and states, compressing long histories into effective working memory, managing structured degradation within set budgets, and normalizing tool outputs for traceability.
- Capability Foundation: Establish foundational systems such as sandbox environments (using Docker or Kubernetes), Memory Store, MCP Hub, Skills Engine, and file system/upload pipelines. Maintain strong multi-tenant isolation, security, and observability across the platform.
Location
This position is based in Suzhou.
