Arthur Vega

Architected Sage’s agentic conversational AI platform using React, FastAPI, LangGraph, tenant-scoped SQL, document retrieval, and sandboxed Python. Implemented tenant-isolated RAG with GCP Discovery Engine and event-driven imports, plus PostgreSQL-backed indexing ledgers. Built an “Agent Maestro” LangGraph authoring pipeline with blueprint validation, tool selection, evaluation preflight, and resumable human approval. Reduced hallucination and unsupported-reasoning failures by hardening routing and repair across RAG and multi-agent workflows. Led Firestore-to-CloudSQL migration with backfills, parity validation, dual-store rollout, rollback controls, and production-readiness gates, and improved retrieval/data quality by tracing cross-tenant invalid records.

AI-powered phone agents that automate healthcare billing and administrative workflows. Led full-stack development of a conversational AI voice-agent (Next.js, NestJS, Python) to automate healthcare calls (scheduling, reminders, insurance checks) using speech recognition/NLP, freeing staff and reducing call-center workload by 50%. Engineered backend telephony services to process inbound/outbound calls and a natural-language dialog engine using LiveKit and LangGraph, enabling AI agent to speak naturally and autonomously handle 50% of routine appointment calls. Integrated with 80+ EHR systems via secure APIs (FHIR/HL7) to fetch live patient data and update records in real-time, enabling personalized scheduling and eligibility scripts, supporting rapid 3-week deployments. Developed an AI Agent Manager dashboard (Next.js/TypeScript) to let non-technical staff customize call scripts and rules; created multi-tenant backends to store client scripts and terminology and translate to unique conve

Built an AI voice-agent platform for healthcare billing and administrative automation, including scheduling, reminders, and insurance verification, reducing call-center workload by 50%. Engineered inbound/outbound telephony and natural-language dialog workflows using LiveKit, LangGraph, and NestJS, enabling 50% of appointment calls to be handled autonomously. Integrated with 80+ EHR systems via FHIR/HL7 for real-time patient data retrieval and updates. Developed a Next.js/TypeScript drag-and-drop Agent Manager for non-technical staff to configure call scripts and business rules. Implemented AI-agent deployment pipelines with Jenkins, Docker, and Terraform across AWS, Azure, and GCP to keep deployment downtime under 5 minutes, and created Grafana dashboards for operational and business metrics.