AI Developer

Cobalt Benefits Group (CBG), a healthcare third-party administrator, seeks an AI Developer to design, build, test, and deploy governed AI solutions that improve operational efficiency and member experience. This hands-on role turns approved use cases into production-ready retrieval-augmented assistants, document-extraction agents, and AI-enabled workflows, using both custom development and low-code Microsoft tools. The developer partners with business owners, DevOps, IT Security, Compliance, and AI governance stakeholders to integrate solutions securely with core systems and establish reusable patterns for responsible AI delivery.

Responsibilities

• Design, prototype, build, test, and deploy AI applications and agents through production; select appropriate Copilot Studio, workflow, agent, custom-service, vendor, or non-AI approaches under Build-vs-Buy guidance.

• Develop retrieval-augmented generation (RAG) solutions, including document ingestion, chunking, indexing, metadata, retrieval tuning, and grounded responses with source citations.

• Integrate with internal and third-party systems through authenticated REST and GraphQL APIs; collaborate on data pipelines for documents and structured data.

• Apply least-privilege permissions, single sign-on, secrets management, PHI/PII protections, output validation, and appropriate logging and retention controls with IT Security.

• Create evaluation harnesses with golden test sets, subject-matter expert rubrics, and production sampling; monitor accuracy, citation support and completeness, latency, and transaction cost, including results by plan type.

• Implement telemetry for prompts, retrieval, tool calls, and outputs; maintain runbooks, hand-off documentation, cost controls, and operational monitoring.

• Advance solutions through CBG AI Solution Delivery Playbook decision gates and required artifacts, including risk assessments, requirements, evaluation plans, and transition documentation.

• Partner with the AI Risk Manager, Compliance, Data Protection Officer, Security, and AI Ethics Committee as appropriate; share reusable components, templates, and approved prompt patterns.