Senior Developer — Agentic Clinical Workflow & Orchestration at SAIGroup

<h2><strong>About the Role</strong></h2> <p>We are developing a next-generation multimodal foundation model capable of understanding&nbsp;<strong>medical imaging</strong>,&nbsp;<strong>clinical documents</strong>, and&nbsp;<strong>operational context</strong>.<br>We are now building&nbsp;<strong>agentic systems</strong>&nbsp;that can execute, coordinate, and optimize&nbsp;<strong>clinical workflows</strong>&nbsp;across radiology, oncology, and cardiology.</p> <p>As a Senior Research Scientist, you will design and implement&nbsp;<strong>workflow-centric AI agents</strong>, including&nbsp;<strong>multi-agent orchestration</strong>,&nbsp;<strong>task routing</strong>, and&nbsp;<strong>clinical workflow automation</strong>.<br>Your work will directly shape how AI collaborates with clinicians and hospital operations teams.</p> <p>&nbsp;</p> <h2><strong>What You Will Work On</strong></h2> <h3><strong>1. Development of Agentic Clinical Workflows (Core Focus)</strong></h3> <ul> <li>Architect AI agents that can execute multi-step clinical workflows end to end, such as: <ul> <li>Imaging ordering → prioritization → acquisition → reporting → follow-up</li> <li>Oncology staging → treatment planning → monitoring → documentation</li> <li>Cardiology testing → diagnosis → risk stratification → longitudinal care</li> </ul> </li> <li>Encode clinical logic, workflow dependencies, safety steps, and escalation patterns into agent behaviors.</li> <li>Build agent pipelines that coordinate across radiology, oncology, cardiology, ED, OPD, and inpatient teams.</li> </ul> <h3><strong>2. Orchestration of Multi-Agent Systems</strong></h3> <ul> <li>Design orchestration frameworks that break down complex clinical tasks into<strong>atomic sub-tasks</strong>&nbsp;and route them to specialized agents (e.g., imaging analysis agent, summarization agent, scheduling agent, retrieval agent).</li> <li>Build controllers that enable agents to share context, verify each other’s outputs, and collaborate dynamically.</li> <li>Implement workflow-aware state machines or controllers to manage long-running, multi-step clinical processes.</li> </ul> <h3><strong>3. Integration With Imaging, Clinical, and Operational Systems</strong></h3> <ul> <li>Develop agents that interact with: <ul> <li><strong>Imaging systems (PACS/RIS)</strong></li> <li><strong>Clinical records (EHR/HIS)</strong></li> <li><strong>Operational systems (scheduling, triage, queue management)</strong></li> </ul> </li> <li>Ensure orchestration layers understand real-world constraints like department availability, urgency, and imaging backlogs.</li> <li>Build agents capable of responding to workflow events (urgent findings, incomplete scans, patient no-shows, etc.).</li> </ul> <h3><strong>4. Leveraging the Multimodal Foundation Model</strong></h3> <ul> <li>Use the foundation model’s imaging-text representations to build agents that can: <ul> <li>Understand 3D studies (CT/MRI/PET)</li> <li>Read clinical notes and operational metadata</li> <li>Perform reasoning across multimodal patient context</li> </ul> </li> <li>Apply prompting, tool use, retrieval augmentation, and memory mechanisms to power agent workflows.</li> </ul> <h3><strong>5. Evaluation, Safety &amp; Reliability</strong></h3> <ul> <li>Define evaluation methods for agentic workflows, including: <ul> <li>Workflow completion accuracy</li> <li>Task routing correctness</li> <li>Operational efficiency improvements</li> <li>Clinical coherence and safety</li> </ul> </li> <li>Collaborate with clinicians to identify edge cases and ensure agents support safe clinical practice.</li> <li>Build monitoring and guardrail mechanisms for workflow-level safety.</li> </ul> <p>&nbsp;</p> <h2><strong>Why This Role Is High-Impact</strong></h2> <ul> <li>You are designing the<strong>workflow brain</strong>&nbsp;of next-generation clinical AI systems.</li> <li>Your work shapes AI’s role in radiology, oncology, and cardiology operational environments.</li> <li>You define new architectures for orchestrated AI that improve clinical efficiency and patient care.</li> <li>You help build the world’s first<strong>agentic clinical coordination layer</strong>&nbsp;on top of a multimodal foundation model.</li> </ul> <p>&nbsp;</p> <h2><strong>What We’re Looking For</strong></h2> <ul> <li>Strong experience in LLM agents, workflow automation, or multi-agent systems.</li> <li>Background in multimodal or clinical AI is a plus.</li> <li>Familiarity with hospital workflows—especially in imaging-heavy specialties—is strongly preferred.</li> <li>Ability to design and implement agent orchestration frameworks.</li> <li>Hands-on experience with PyTorch/JAX, transformers, retrieval systems, or agent frameworks (AutoGen, CrewAI, LangChain, etc.).</li> <li>Ability to collaborate with clinicians and product teams to translate real-world workflow needs into agentic systems.</li> </ul> <p>&nbsp;</p> <h2><strong>Nice to Have</strong></h2> <ul> <li>Experience with PACS/RIS/EHR integration, clinical informatics, or operational workflow tools.</li> <li>Knowledge of oncology and cardiology care pathways.</li> <li>Prior work with workflow engines, BPMN systems, or orchestration controllers.</li> <li>Experience running simulations, RL-based workflow optimizers, or task planners.</li> </ul> <p>&nbsp;</p> <h2><strong>What We Offer</strong></h2> <ul> <li>Competitive compensation commensurate with productivity and innovation.</li> <li>Opportunity to define the orchestration layer for a new generation of clinical AI.</li> <li>High-compute resources.</li> <li>Close collaboration with clinicians, operations leaders, and AI researchers.</li> <li>Freedom to publish and innovate in an emerging field.</li> </ul> <p>&nbsp;</p>