Professional Experience

Co-led architecture for 'The Lennar Machine' dynamic AI pricing platform (Next.js, Prisma, AWS) with CI/CD via Azure Pipelines and GitHub Actions, driving a 1% pricing uplift (~$36M annualized). Stood up Lennar Connect's GenAI platform using LangGraph multi-agent workflows and GPT-4o on Azure to automate Tier-1 HR operations across PTO, ServiceNow, and Workday with governed RAG. Led a 9-engineer team, owned evaluation + production telemetry, and partnered with the CTO on AI roadmap priorities, success metrics, and rollout strategy.

Contributed to Amazon Compliance Screening's AI-powered investigation tier for sanctions screening at Amazon scale (~2B transactions/day), implementing multi-agent workflow patterns and tool interfaces. Implemented deterministic orchestration and dependency sequencing for graph-style agent execution with exception handling, while enforcing SOP compliance through confidence-based human escalation. Built and standardized MCP servers and agent tool integrations for data aggregation, geospatial/address validation, and OSINT lookups, and led 16+ ProServe engagements from discovery through production with VP/Director stakeholders.

Delivered production LLM agent systems using LangGraph with Azure OpenAI and Azure AI Search, including evaluation loops with golden datasets, LLM-judge scoring, and regression gates that track quality, p50/p95 latency, and cost across deployments. Built React Native + Node/AWS event-driven platforms with real-time ranking features, improved startup latency by ~7.5x, and shipped AI search capabilities using Pinecone and Bedrock.

Led end-to-end delivery of ML decisioning systems for fraud, credit risk, and workflow automation in regulated environments, including production rollout, incident response, and performance optimization. Built MLOps foundations for dataset/model versioning, CI/CD for model artifacts, drift monitoring, retraining, and rollback strategies, and operationalized evaluation/risk tradeoffs through release gates.