• 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

• Design and implement API gateways, model routing layers, and service infrastructure that connect fine-tuned CRE models with production applications. • Develop internal developer tooling - prompt management interfaces, evaluation dashboards, trace viewers - that accelerate the team's ability to ship and iterate on AI features. • Implement integration patterns for embedding AI capabilities into LightBox's existing workflow applications, ensuring seamless user experiences. • Build observability and monitoring infrastructure for AI workloads: latency tracking, quality metrics, cost attribution, and usage analytics. • Contribute to the platform's knowledge retrieval and caching layers, optimizing for both accuracy and performance at scale. • Collaborate with ML engineers on model serving infrastructure, helping bridge the gap between training outputs and production deployment. • Participate in architecture decisions and technical design reviews, bringing a pragmatic full-stack perspective to the team's work.

• Design, implement, and maintain the agent harness • Build the systems we use to run and reproduce experiments • Own environments, tooling, and repeatable runs • Partner with security and evaluation engineers on their workflows

• Own the architecture and long-term technical strategy for RapidFort's Build Platform. • Design and operate highly scalable distributed build systems capable of producing and maintaining thousands of production-ready container images. • Build automated systems for image creation, validation, signing, publishing, and lifecycle management. • Improve build performance, scalability, reproducibility, reliability, and operational efficiency. • Design automation that eliminates manual engineering work while improving developer productivity. • Develop internal tooling and platform services using Python, Bash, and modern cloud-native technologies. • Solve complex Linux, dependency management, networking, container, and build-system challenges. • Drive engineering standards around software supply chain integrity, reproducible builds, and platform reliability. • Partner closely with Platform Engineering, Security Research, Product Engineering, and Infrastructure teams. • Lead architecture reviews and influence technical decisions across multiple engineering organizations. • Mentor senior engineers and help raise the technical bar across the company. • Continuously identify opportunities to simplify systems, improve reliability, and reduce operational overhead.