(Principal) Scientist, Computing & Intelligence, IHPC

Job Description

The Senior AI Research Scientist will be working on clinical multi-agent CDS engineering workstream. The role will design and implement the MAS harness that coordinates GP, pharmacist, specialist, guideline/evidence, medication-safety and workflow agents for primary-care decision support. The system should support pre-consult briefs, iterative re-inference during consultation, safety checks, and structured next-best-action recommendations.

The scientist will be responsible for translating clinical workflow requirements into an executable architecture: agent contracts, orchestration logic, memory/state management, tool interfaces, retrieval layers, evaluation harnesses and observability. The role will also lead simulation-based testing and validation using clinical vignettes, synthetic cases, public benchmarks and protected clinical data pathways where approved.

Key Responsibilities

• Design and implement the MAS/CDS harness, including agent roles, orchestration policies, memory/state handling, tool calling, guardrails and error recovery.
• Build simulation-based evaluation workflows that generate cases, run multi-agent consultations, compare outputs with labels or clinician review, and record failure modes.
• Engineer validation pipelines for guideline compliance, medication safety, role adherence, hallucination detection, uncertainty handling, and workflow usability.
• Integrate clinical foundation models, retrieval components, guideline knowledge bases, structured patient data and clinician feedback loops into a coherent CDS prototype.
• Lead technical design for pre-consult briefs and iterative consultation support, including re-inference when new patient information, lab results or clinician inputs are added.
• Collaborate with clinicians, product/workflow teams and evaluation teams to define acceptance criteria, benchmark scenarios, safety thresholds and pilot-readiness evidence.
• Mentor junior researchers/engineers and establish engineering standards for reproducible MAS experiments, audit trails, dataset versioning and model/system documentation.

Required Qualifications and Skills

• PhD degree in AI, computer science, machine learning, biomedical informatics, computational science or a related field.
• Strong hands-on experience building LLM applications, agentic systems, orchestration frameworks, evaluation harnesses or production-grade AI research prototypes.
• Deep understanding of LLM evaluation, retrieval-augmented generation, tool use, safety guardrails, observability, state management and experiment reproducibility.
• Ability to design validation approaches for clinical AI systems, including synthetic and real-world data evaluation, clinician review workflows and error analysis.
• Strong software engineering skills in Python and modern AI system stacks; able to convert research ideas into maintainable prototypes and reusable platforms.

Preferred Experience

• Experience with clinical decision support systems, medical LLMs, healthcare workflow integration, FHIR/EMR/eHINTS-like data pathways or regulated AI evaluation.
• Experience with multi-agent frameworks, simulation environments, LLM-as-judge systems, benchmark construction, or safety testing for high-stakes AI.
• Knowledge of primary-care chronic disease management and clinical safety issues such as polypharmacy, contraindications, formulary constraints and care-gap detection.
• Track record leading small technical teams, mentoring junior researchers, and coordinating with clinical, product and governance stakeholders.