Research Fellow (AI / Computer Science / Machine Learning / NLP)

About the role

ATMRI at NTU is seeking a Research Fellow to join the Digital ATCO Assistant project. The role leads the assistant's knowledge, retrieval, reasoning and agent orchestration capabilities through LLMs, Agentic AI, multi-agent architectures and Retrieval-Augmented Generation.

Key responsibilities

• Design, develop and evaluate Agentic AI and multi-agent architectures for task decomposition, planning, reasoning, coordination and workflow orchestration.
• Develop tool and function calling so agents can select and invoke tools, APIs and knowledge services, including input validation, output interpretation and error handling.
• Build the retrieval agent and end-to-end RAG pipelines across AIPs, NOTAMs, operational procedures and regulations using vector databases, semantic retrieval and knowledge graphs.
• Design benchmarks and run experiments to evaluate agent performance, tool-call correctness, retrieval precision and recall, factuality, grounding and hallucination.
• Work with the integration engineer/technical lead to expose agent, retrieval and reasoning capabilities via APIs and microservices for the prototype.
• Work with domain experts and controllers to turn operational use cases into research requirements, evaluation datasets and test scenarios.
• Contribute to technical reports, project proposals, publications, demonstrations and collaboration with academic, industry and regulatory partners.

Required profile

• PhD in AI, Computer Science, Machine Learning, NLP or a related discipline, or a Master's degree with substantial applied research experience.
• 2-4 years of hands-on experience with LLMs, Agentic AI, multi-agent architectures, tool/function calling and prompt/context engineering.
• Experience with RAG, vector databases, semantic retrieval, knowledge graphs, grounded generation and hallucination mitigation.
• Good communication and multidisciplinary collaboration skills; peer-reviewed publications and experience in safety-critical domains are preferred.

Required skills

• Strong Python and PyTorch, modern LLM frameworks, APIs and microservices.
• Evaluation of retrieval precision/recall and factuality.
• LoRA/PEFT, in-context learning, Graph RAG, temporal RAG, hybrid symbolic-neural reasoning, on-premise deployment or inference optimisation are highly desirable.