Agent Application Development - LLM Track
Description
About the role
We are looking for an engineer specialized in LLM Agent systems. You will own the design, prototyping and production deployment of autonomous agents, multi-agent workflows, tool calling pipelines and grounded RAG systems.
Key responsibilities
Design and implement LLM agent architectures: task planning, tool use, long-term memory, state management and human-in-the-loop workflows with frameworks such as LangGraph, AutoGen, CrewAI
Build and optimize RAG pipelines: document parsing, vector retrieval, grounding, hallucination mitigation and evaluation metrics
Conduct prompt engineering, supervised fine-tuning, and develop automated benchmarks for agent capability (GAIA, SWE-bench style evaluation)
Build backend services, agent sandbox environment, API orchestration, deployment monitoring and failure tracing
Track cutting-edge agent research (reflection, multi-modal agents, A2A protocol), run PoCs and deliver technical reports
Collaborate with the AI evaluation team to design agent test suites and automated assessment pipelines
About you
PhD in Computer Science, AI, NLP or Machine Learning preferred
Publications in top conferences (NeurIPS, ICLR, ACL, ICML) are a big plus
Solid theoretical foundation in LLMs, NLP, machine learning, alignment and context engineering
Proficient in Python; hands-on production experience building agents with LangGraph, AutoGen, LlamaIndex and vector databases (Milvus, Pinecone, Chroma)
Proven experience in tool-calling, multi-agent collaboration and end-to-end RAG systems; prior work on agent evaluation benchmarks is highly desirable
Good English communication, able to read research papers and write technical documentation
Work onsite at Cyberport, Hong Kong; familiar with Hong Kong tech ecosystem is a bonus
Benefits
Competitive compensation and performance bonus
Flexible working at Cyberport
Work on cutting-edge agentic AI and AI evaluation products
Visa sponsorship support for eligible candidates