MVA LLM Developer 奔驰虚拟助手开发工程师

Design and develop the next-generation in-car MVA Agent architecture,including:

• Build agentic systems with design patterns such as Re Act, Plan-and-Execute, and Tool Calling / Function Calling
• Integrate LLM provider APIs and implement multi-model orchestration and routing strategies
• Develop conversational memory services covering short-term dialogue context and long-term user profiling
• Design and implement standardized tool-calling interfaces for vehicle capabilities
• Perform Prompt Engineering, RAG pipeline development, and model evaluation to continuously optimize AI quality
• Conduct unit testing, integration testing, and AI-specific evaluation (eval datasets, A/B testing) to ensure delivery quality
• Submit and deploy code via CI/CD pipelines on Azure cloud

Drive cross-team technical alignment and solution design as the technical owner for the Agent architecture

Produce and maintain system architecture proposals, interface design documents, and technology selection reports, including:

• Output architecture proposals for new Agent features and review them with the team lead
• Maintain API specifications and tool-calling schemas
• Document AI system behavior, prompt versioning, and evaluation results across the product lifecycle

Continuous improvement of AI development methodologies and tool chains including:

• Evaluate and adopt new LLM capabilities, Agentf rameworks, and AI coding tools (Cursor, Copilot, Claude Code)

• Monitor production performance, track AI quality metrics, and drive post-go-live optimization

• Stay current with industry trends in LLM, Agent,and automotive AI

• Bachelor's degree or above in Computer Science, Software Engineering, Artificial Intelligence, or a related field

• At least 3+ years of experience in LLM application development, AI Agent development, or similar AI engineering roles

• Hands-on experience in building AI Agent systems from scratch; familiar with agentic design patterns such as Re Act, Plan-and-Execute, Tool Calling / Function Calling, and agent memory management

• Familiar with at least one mainstream agent orchestration framework (e.g. Lang Chain / Lang Graph, Llama Index, Auto Gen, CrewAI, Semantic Kernel, or Dify / Coze)

• Deep understanding of mainstream LLM capabilities and limitations (e.g. GPT-5, Claude, Doubao/Seed, Qwen, Deep Seek, GLM, Mini Max); experience with multi-model orchestration and routing is a plus

• Proficient in Prompt Engineering: System Prompt design and optimization, Chain-of-Thought (CoT), Few-shot prompting, Prompt Chaining, structured output generation (JSON mode / Function Calling), and guardrails

• Proven project experience with RAG: hands-on with vector databases (Milvus, Pinecone, Chroma, pgvector, or similar), embedding strategies, chunking strategies, hybrid search (semantic + keyword), and re-ranking

• Experience in AI system evaluation: evaluation metric design, A/B testing for AI features, building evaluation pipelines (e.g. RAGAS or custom eval frameworks)

• Understanding of conversational memory architectures: short-term dialogue context management and long-term user profiling / episodic memory

• Strong Python programming ability (primary language for AI development); Go language development experience (team's main tech stack)

• Proficient in cloud technology, especially micro-service architecture, API design, and cloud-native deployment on platforms such as Azure

• Ability to independently produce system architecture proposals, interface design documents, and technology selection reports

• Proficient in AI-assisted programming tools such as Cursor, GitHub Copilot, or Claude Code for daily development

• Strong ownership mentality — proactively drives cross-team alignment without waiting for instructions

• Self-driven, team-oriented, and excellent communication skills; able to simplify complex problems and decompose them into actionable tasks

Good English communication skills, Good German language is a plus