Job Description
Backend Development Engineer - AI Agent Direction
About This Role:
We are seeking a backend engineer who embraces the "Agent Native" philosophy. Unlike traditional CRUD development, you will be responsible for designing and building intelligent agent systems capable of autonomous planning, memory retention, tool invocation, and environmental interaction.
We don’t expect you to treat AI as just an occasional API call. Instead, you will reconstruct backend logic using state machines, graph computation (Graph), and autonomous decision-making as core paradigms. Here, every line of code you write may determine how an Agent understands complex tasks and breaks them down for execution.
Key Responsibilities
- Core Framework Development: Design and implement a highly scalable AI Agent execution engine supporting various reasoning modes such as ReAct, Plan-and-Execute, and multi-agent collaboration.
- Tool & Ecosystem Integration: Develop "hands and feet" for Agents—seamlessly encapsulate internal APIs, third-party services, and databases into standardized tools via Function Calling/Tool Use mechanisms.
- Memory System Construction: Design hybrid memory architectures, including long-term memory (vector database-based) and short-term working memory (Redis-based).
- Workflow Orchestration: Use LangGraph, DSPy, or custom DSL (Domain-Specific Language) to orchestrate complex Agent workflows, handling loops, retries, backtracking, and human-in-the-loop collaboration.
- Performance Optimization: Optimize LLM call latency and costs (caching, prompt compression, model routing) while ensuring the stability of asynchronous task queues under high concurrency.
Job Requirements
- Education & Experience: Computer-related major with 1-2 years of AI Agent development experience.
- Solid Backend Foundation:
- Proficiency in at least one of Python/Go/Java (Python preferred due to its mature AI ecosystem).
- Mastery of asynchronous programming (e.g., Python asyncio, Go Goroutine).
- Familiarity with frameworks like FastAPI/Spring Boot and ability to independently design RESTful/gRPC APIs.
- Databases & Middleware: Proficient in PostgreSQL/MySQL and familiar with at least one vector database (Milvus/Pinecone/Qdrant/Chroma) for basic use cases (no tuning required, but understanding of vector retrieval is essential).
- AI Fundamentals:
- Deep understanding of LLM limitations (hallucinations, context windows, reasoning bottlenecks) and ability to mitigate them through engineering solutions.
- Practical experience in Prompt Engineering (Few-shot, Chain-of-Thought).
Benefits
- Team-building activities
- Health check-ups
- Year-end bonuses
