As Agentic AI Engineer, you will design, build, and operationalize agentic AI capabilities to increase automation, improve data quality, and accelerate decision-making within medical imaging clinical trials. This role focuses on building AI agents that can reliably execute workflows such as intake and validation of imaging data, metadata reconciliation, protocol-driven checks, quality control support, and structured extraction from multi-modal artifacts (DICOM, non‑DICOM, reports, PDFs).  In parallel, the role will help transform the company’s Product Development Life Cycle (PDLC) into an AI-driven lifecycle—enabling agentic workflows from Product Discovery (e.g., requirements synthesis, research, backlog shaping) to Delivery and Deployment (e.g., test generation, release readiness, documentation automation) through Production Operations (e.g., incident triage, observability insights, automated runbooks), while ensuring safety, compliance, traceability, and human oversight.

Key Responsibilities

Cross functional collaborations

• Partner with Product, Imaging Ops, Data Engineering, QA, Security, and Regulatory/Quality to identify high-value automation opportunities and define agentic AI use cases.
• Translate clinical-trial imaging workflows and PDLC processes into agent-ready task models (inputs, outputs, constraints, guardrails, acceptance criteria).
• Work with SMEs to define quality rubrics, evaluation datasets, and human review workflows to validate agent outputs.
• Support change management: help teams adopt agentic features with clear UX patterns, onboarding, and documentation.

Technical ownership

• Promote engineering best practices for agentic systems: reliability, observability, evaluation, reproducibility, and secure-by-design patterns.
• Contribute to internal standards for: prompt/version control, agent tools/plugins, retrieval configuration, grounding/citations, and human-in-the-loop review.
• Participate in design reviews and provide guidance on selecting frameworks and architectures.
• Coach peers through code reviews, pairing, demos, and knowledge-sharing sessions.

Algorithm and Software Development

• Build agentic workflows (multi-step reasoning + tool use) that support imaging trial processes such as:
• data intake and validation (DICOM/non‑DICOM), metadata consistency checks
• protocol-driven QC support and discrepancy detection
• data completeness checks and structured issue creation
• automated summarization of imaging packets and context generation for reviewers
• Implement retrieval-augmented generation (RAG) patterns grounded in controlled knowledge sources (SOPs, protocols, specs, runbooks), with traceability.
• Develop evaluation harnesses for agent behavior: task success rate, hallucination/grounding checks, regression testing, and bias/failure-mode analysis.
• Contribute production-quality code (APIs, services, pipelines) with strong testing, documentation, and maintainability.

Deployment and integration (MLOps)

• Package and deploy agentic services using approved cloud and SDLC patterns: CI/CD, infrastructure-as-code, secret management, and environment promotion.
• Implement guardrails and controls: access control, tool permissions, sandboxing, content filtering, rate limits, and audit logging.
• Build observability: traces, structured logs, evaluation telemetry, cost monitoring, and error analytics (agent runs, tool calls, retrieval hits, model performance).
• Ensure safe production operations: rollback strategies, feature flags, canary releases, and incident playbooks.

Innovation & Research application

• Stay current on agentic AI patterns (planning, tool use, memory, multi-agent coordination, structured outputs) and evaluate feasibility for regulated workflows.
• Prototype and validate new capabilities quickly, then harden into production-ready features.
• Run experiments comparing models, prompts, retrieval strategies, and agent architectures; document results and recommendations.
• Identify opportunities to standardize and reuse agent components across clinical imaging workflows and PDLC automation.

Other

• Carry out any other reasonable duties as requested.
•