AI Engineer, Agent/Platform Tracks - Parexel

We are seeking an AI Engineer, Agent/Platform Tracks to join our team. As an AI Engineer focused on agent development, you will drive the design and implementation of specialized pharmacovigilance agents that transform how adverse event data is processed and analyzed. This position plays a critical role in building intelligent systems that advance clinical research safety and efficiency, working alongside domain experts to create cutting-edge AI solutions that set new standards in the industry. You'll join a fast-paced, innovation-driven environment focused on making a meaningful impact through AI-powered pharmacovigilance automation. With diverse teams and continuous learning opportunities, this role offers pathways to deepen your expertise in large language models, prompt engineering, and healthcare AI while influencing the future of clinical research safety systems. Key Responsibilities Implement the specialized pharmacovigilance agents: write system prompts, configure model parameters, build tool-use definitions, and define agent boundaries to ensure precise adverse event processing Build and iterate prompt chains for each processing step: source document parsing, field extraction, MedDRA coding suggestions, causality assessment logic, narrative drafting, and E2B(R3) output generation Develop the deterministic rule engine layer: implement ICH E2B field validation checks, MedDRA hierarchy verification, and regulatory logic constraints that operate alongside LLM outputs Create and maintain evaluation datasets in collaboration with the pharmacovigilance domain team: annotated ground-truth cases, edge case libraries, and regression test suites Develop and maintain Model Context Protocol (MCP) servers to expose enterprise applications, APIs, databases, and services as standardized tools for AI agents Implement secure MCP integrations, tool definitions, authentication, and testing to enable reliable agent interaction with internal and external systems Run accuracy benchmarks, analyze failure modes, and iterate on prompts and agent configurations to improve performance against defined thresholds Implement the quality control agent's cross-verification logic: configure separate Claude instances, build comparison algorithms, and calibrate confidence scoring Build human-in-the-loop feedback mechanisms: reviewer interfaces for accept/modify/reject decisions, structured feedback capture, and feedback-to-prompt-improvement pipelines You'll thrive in this role if you bring: Strong prompt engineering skills and experience writing and iterating system prompts, few-shot examples, chain-of-thought patterns, and structured output formats Solid foundation in Python with hands-on experience in LLM orchestration frameworks such as LangChain, LangGraph, or similar tools .