Senior AI Agent Platform Engineer

Our client, a top-tier crypto exchange with global business layout, is currently making heavy strategic investments to integrate AI technology with core business scenarios and build next-gen intelligent business infrastructure. To support the rapid iteration and commercial implementation of its AI ecosystem, we are recruiting a senior technical expert to join the core team as a Senior AI Agent Platform Engineer . The team is building a production-grade, enterprise-level AI Agent platform that empowers customizable, adaptive, and verifiable intelligent agents for full-scenario business automation.

We’re looking for a senior backend/AI engineering expert who has end-to-end 0-to-1 Agent platform building & iteration experience to drive the continuous iteration and large-scale commercial landing of the company’s Agent infrastructure.

Core Job Responsibilities

1. Agent SDK & Runtime Infrastructure

Design and develop core Agent orchestration engines, runtime environments and underlying primitives. Build a controllable, verifiable agent execution loop while ensuring natural conversational interaction and dynamic adaptive capabilities.

2. MCP System Capability Expansion

Encapsulate internal business systems into standardized MCP servers, sub-agents and reusable agent skills. Build unified management capabilities including tool discovery, dynamic registration, RBAC permission control, traffic rate limiting and version iteration.

3. Full Lifecycle Agent Engineering (ADLC)

Tailor dedicated toolchains for the non-deterministic characteristics of AI agents. Cover the entire closed-loop workflow: requirement specification & evaluation → development & testing → A/B testing & canary release → real-time monitoring → continuous iteration.

4. End-to-end Business & Cost Ownership

Take full ownership of core platform metrics including accuracy, problem resolution rate, service latency and resource cost. Build data-driven agent personalized optimization loops to balance user experience and operational cost.

Job Requirements

• Bachelor’s degree or above in Computer Science, Software Engineering or related majors; - P7: 6+ years of engineering experience, including 3+ years in AI/Agent field; - P7+: 8+ years of engineering experience, with proven 0-to-1 Agent platform independent construction experience .
• Proficient in Go or Python; capable of independent production-level system design and complete TRD technical document output.
• Deep practical experience in LLM API docking, prompt engineering, ReAct agent patterns, and core context management (token budgeting, conversation history compression, sub-agent context optimization).
• Complete experience in 0-to-1 launch and continuous iteration of commercial Agent products (excluding simple Demo/POC prototype projects).
• Familiar with mainstream orchestration frameworks (LangGraph / Dify / Eino / OpenAI Agents SDK, etc.); have hands-on experience in multi-agent collaboration scenarios including team collaboration, parallel execution, DAG scheduling and task handoff.
• Proficient in one of the two core directions: RAG engineering (vector retrieval, hybrid search, reranking optimization) . Agent evaluation system (LLM-as-Judge, golden dataset construction, regression testing, A/B experimental analysis)
• Able to quantitatively optimize model effect and token cost; familiar with PII data desensitization and prompt injection defense; understand compliance boundaries in financial/business scenarios.

Preferred Qualifications (Bonus Points)

• 0-to-1 experience in building Agent simulation & benchmarking platforms (τ-Bench equivalent scale)
• Private LLM deployment experience (vLLM / Triton / DeepSeek / Qwen / GLM)
• Domain expertise in Crypto AI (risk control / KYC / compliance), AIOps, customer service AI, Code AI
• AI security research achievements (prompt injection defense, jailbreak resistance, smart contract AI security)
• Open-source contributions to LangChain / LangGraph / Dify / MCP / Eino / Vercel AI SDK and other mainstream agent ecosystems
• Public technical influence (tech blogs, conference speeches, academic papers, open-source projects)