Kevin Cassani

Developed autonomous clinical reasoning agents with Python using LangGraph/LangChain to automate medical chart reviews for Optum Integrity One, reducing administrative documentation time by 30% for 5,000+ providers. Built and scaled a production RAG system using Azure OpenAI (GPT-4o) to process InterQual 2025 guidelines, achieving 94% accuracy for evidence-based clinical recommendations. Architected real-time clinical dashboards with React 19 and Tailwind/shadcn, using TanStack Query to reduce claim adjudication latency to <300ms. Integrated a proprietary clinical language intelligence engine into FastAPI backends for autonomous extraction and coding of unstructured medical data at 92% precision. Deployed enterprise-grade AI APIs for an AI Marketplace with Pydantic v2 validation for FHIR/HL7 and HIPAA compliance. Managed containerized GenAI microservices on Azure Kubernetes Service using Docker and Terraform to maintain 99.99% uptime during national rollout, and optimized vector simila

Developing autonomous clinical reasoning agents using Python (LangGraph, LangChain) to automate medical chart reviews for Optum Integrity One, reducing administrative documentation time by 30% for 5,000+ providers. Built systems that integrate external APIs and use LLMs for concise, actionable summarization. Architected and productionized a RAG solution over InterQual 2025 guidelines using Azure OpenAI (GPT-4o) with 94% accuracy for evidence-based clinical recommendations. Built real-time clinical dashboards using React and Tailwind, optimizing server-state caching to reduce latency to under 300ms. Integrated Optum Clinical Language Intelligence™ engine into FastAPI backends for autonomous extraction and coding of unstructured medical data with 92% precision. Developed and deployed HIPAA-compliant enterprise APIs for the Optum AI Marketplace using Pydantic v2 for strict FHIR/HL7 validation. Managed containerized GenAI microservices on Azure Kubernetes Service with Docker and Terrafor

Engineered a RAG pipeline using Python (FastAPI) and watsonx.ai with Granite 3.0 LLMs, enabling financial clients to query complex regulatory data with 95% accuracy. Developed 8+ autonomous digital labor agents with watsonx Orchestrate to automate multi-step API integrations between React front ends and Python microservices, contributing to a company-wide reduction of 3.9 million work hours in 2024. Built a prompt-engineering library that automated PyTest/Jest generation and refactoring to accelerate feature turnaround time by 30%. Architected high-concurrency dashboards using React and Carbon Design System, achieving 100% WCAG 2.2 accessibility compliance. Orchestrated migration from monolithic Python applications to containerized microservices on OpenShift with a 99% success rate. Provisioned secure hybrid cloud infrastructure using Terraform and implemented AI model safety governance with watsonx for privacy in regulated client environments.