Multi-agent collaborative engineer

Job Title: Multi-Agent Collaborative Engineer
Job Responsibilities

• Design and implement multi-business Agent's role division and responsibility boundary, covering pricing, inventory, fulfillment, channel, planning, etc. core business modules

• Build an Agent life cycle management mechanism to support dynamic behaviors such as startup, pause, recovery, and restructuring, ensuring system flexibility and stability

• Define Agent authority boundaries and separation of responsibilities, ensuring module autonomy and system security

• Develop agent-to-agent communication protocol and state broadcasting mechanism, realizing cross-agent state synchronization and data consistency

• Build a conflict detection and negotiation framework, handle business goal conflicts such as "price vs inventory" and "lead time vs cost", and support learnable negotiation decisions

• Integrate centralized and decentralized hybrid decision-making architectures, and introduce multi-agent reinforcement learning (MARL) technology to continuously optimize overall system benefits

• Build a multi-agent simulation and replay environment to support strategy pre-verification and anomaly reproduction

• Build a collaborative behavior observability system, covering decision path tracking, responsibility attribution analysis, and effect quantification evaluation

• Set up an A/B testing framework for scientifically evaluating the actual impact of different collaborative strategies

• Design an open access protocol to support new business modules (such as promotion Agent) quickly integrate into existing collaborative networks

Work requirements

• Bachelor's degree, software engineering, computer science, artificial intelligence, automation, mathematics or related fields preferred

• Familiar with Python, with solid programming design skills and engineering practice experience; those with PyTorch usage experience are preferred.

• Preference will be given to those with experience in developing complex systems, distributed systems or agent (Agent) systems

• Understand the basic concepts of reinforcement learning or game theory, and be able to translate theoretical tools into practical collaborative mechanisms

• Have the ability to decompose complex business logic into multiple autonomous modules, and be able to perform system-level abstraction and architecture design

• Have sharp insights into system degradation, goal conflicts, and strategy drift in real industrial scenes, and be willing to deeply participate in solving

• Able to communicate in Mandarin and Cantonese on a daily basis; have good written expression and cross-team collaboration skills

• No experience restrictions, interns and fresh graduates are welcome to apply; those with any of the following backgrounds will be given priority:
  — Multi-agent reinforcement learning (MARL) project experience
  — Game theory or mechanism design related research/practice
  — Scheduling system, resource allocation system development experience
  — Enterprise-level distributed systems or highly available service architecture experience

Benefits

• Compensation

• Year-end bonuses

• Holiday for marriage