(Suitable candidates has to relocate to Hong Kong, KL Malaysia or Abu Dhabi)

Key Responsibilities

• Agent SDK & Runtime — Design and implement the orchestration engine, runtime, and core primitives; ensure the agentic loop is both steerable and verifiable while remaining conversational and adaptive.
• MCP Matrix Expansion — Wrap internal systems as MCP servers / sub-agents / agent skills; unify tool discovery, dynamic registration, RBAC, rate limiting, and version management.
• Agent Development Lifecycle (ADLC) — Build toolchains adapted to the non-deterministic behavior of agents, covering the full loop: Spec → Eval → Build → Test → A/B → Canary → Monitoring → Iteration.
• End-to-End Ownership of Effectiveness & Cost — Own accuracy, resolution rate, cost, and latency metrics; build agent personalization optimization loops through data feedback.

Requirements

• Education & Experience: Bachelor’s degree or above in CS/SE or related field; 6+ years of engineering experience with 3+ years focused on AI/Agent systems; candidates with 8+ years and hands-on 0-to-1 Agent platform delivery are strongly preferred.
• Engineering Foundation: Proficient in Go or Python; capable of production-grade system design and TRD documentation.
• LLM Application Engineering: Deep expertise in LLM APIs, Prompt Engineering, ReAct patterns, and context window management (token budgeting / history compression / sub-agent compression).
• Production-Grade Agent Experience: Full 0-to-1 launch with continuous iteration loop — clearly distinguished from Demo/POC work.
• Agent Engineering Practice: Familiar with at least one orchestration framework (LangGraph / Dify / Eino / Embabel / OpenAI Agents SDK, etc.); hands-on experience with Multi-Agent collaboration (Agent Teams / parallel / DAG / Handoff).
• Adjacent Agent Capabilities: Deep expertise in at least one of: RAG engineering (vector retrieval, Hybrid Search, Reranking) or evaluation systems (LLM-as-Judge, Golden Set, regression testing, A/B).
• Engineering Mindset: Ability to quantify effectiveness and cost (token optimization thinking); understanding of PII desensitization and prompt injection protection; ability to identify compliance boundaries in financial scenarios.

Bonus Points

• 0 to-1 experience building an Agent simulation / benchmarking platform (τ-Bench scale)
• Private LLM deployment: vLLM / Triton / DeepSeek / Qwen / GLM
• Domain expertise: Crypto AI (risk control / KYC / compliance), AIOps (alert RCA / diagnostic tooling), customer service AI, code AI
• AI security research: published work on prompt injection, jailbreak defense, Smart Contract AI security, etc.
• Open-source contributions to LangChain / LangGraph / Dify / MCP / Eino / Vercel AI SDK or similar
• Public technical influence: blog posts / conference talks / papers / OSS projects