Location
Singapore
Department
Institute of High Performance Computing
Posted
1 week before
About This Role
This role involves designing and implementing a Multi-Agent System (MAS) harness for clinical decision support in primary care. The scientist translates complex clinical workflows into an executable, validated AI architecture.
Responsibilities
- Design and implement the MAS/CDS harness, including agent roles, orchestration, and memory management.
- Build simulation-based evaluation workflows for multi-agent consultations and failure mode recording.
- Engineer validation pipelines covering guideline compliance, safety, and hallucination detection.
- Integrate clinical foundation models, retrieval components, and structured patient data into the CDS prototype.
- Lead technical design for pre-consult briefs and iterative consultation support during patient care.
- Collaborate with clinical teams to define acceptance criteria, safety thresholds, and pilot readiness.
Requirements
- PhD in a relevant technical field (AI, CS, ML, etc.).
- Hands-on experience with LLM applications and agentic systems.
- Deep understanding of LLM evaluation, RAG, guardrails, and state management.
- Ability to design validation approaches for clinical AI using synthetic/real data.
- Strong software engineering skills in Python and modern AI stacks.
- Proven ability to convert research ideas into maintainable prototypes.