Build advanced Agent architectures, LLM protocol integrations, AI-native development workflows, and production AI services.
Responsibilities
- Design and implement AI Agent systems with complex reasoning, multi-step planning, and self-correction using patterns such as ReAct, Reflexion, and CoT.
- Research and integrate mainstream protocols such as OpenAI, Anthropic, and MCP to optimize model invocation, context utilization, and inference stability.
- Apply AI-native development workflows by expressing high-level architectural intent and driving rapid implementation.
- Turn advanced AI research into scalable production services with reliable, low-latency responses under high concurrency.
Requirements
- Deep understanding of Transformer architecture, attention mechanisms, KV Cache optimization, quantization, and tokenization effects.
- Strong knowledge of low-level LLM interaction protocols, including streaming, state management, and function-calling edge cases.
- Expertise in end-to-end RAG optimization, including chunking, reranking, and multimodal data integration.
- Fluency with modern AI development tools such as Claude Code, Cursor, and Codex, plus strong code taste and architecture judgment.
- Ability to build and refactor systems through AI-native workflows driven by intent rather than manual coding alone.
- Excellent async programming and systems-architecture skills in any strong language such as Python, Rust, Go, or Node.js.
- Experience with vector databases, graph databases, and high-performance caching.
- Ability to design and run evaluation frameworks for AI accuracy and safety.
Nice to have
- Experience building distributed AI systems, multi-agent collaboration networks, or complex autonomous workflows.
- Deep knowledge of LLM infrastructure such as local inference acceleration or fine-tuning.
- Active GitHub or technical-community presence with notable AI projects or strong technical writing.
What we offer
- Full reimbursement for mainstream AI development tools, IDEs, and model API subscriptions.
- Flat team structure that supports fast AI-native product iteration.
- Work close to frontier AI technologies and protocols while shaping next-generation AI applications.