Tony Hunter

Built a multi-tenant healthcare AI agent platform answering patient calls and texts to support appointment booking and billing questions. Designed real-time voice pipelines using Twilio and Deepgram (streaming speech-to-text, LLM inference, and text-to-speech) across scheduling, intake, billing, and administrative workflows. Implemented LangGraph orchestration with tool calling and structured state, including output validation, tool authorization, confidence-based escalation, and human handoff to prevent unsupported answers and unauthorized patient-data actions. Developed async FastAPI services with PostgreSQL, Redis, and SQS using idempotency, retries, and dead-letter queues, and added secure tenant-authenticated APIs with audit logs. Set up AWS infrastructure and observability (ECS/Lambda/RDS, Docker, Terraform/CDK, OpenTelemetry/Langfuse/CloudWatch), and built React/Next.js admin tooling for tenant configuration, call review, transcripts, and operational monitoring.

Created AI-assisted documentation workflows that turn doctor-patient conversations into draft medical notes for clinician review. Built Python/FastAPI/Node.js services and PostgreSQL-backed workflows for audio ingestion, transcription orchestration, processing state tracking, and delivery of structured results to documentation systems. Developed NLP extraction that identifies symptoms, medications, diagnoses, and clinical context from transcripts and links generated notes back to source segments for review. Split ingestion, transcription, NLP, and note generation across workers using SQS/Kafka/Redis, and integrated with EHR systems using HL7/FHIR where required. Shipped clinician review tools in React/TypeScript with encryption, access controls, audit trails, and evaluation datasets.

Developed microservices for member enrollment, eligibility, and claims administration in a health insurance environment. Implemented Python/Java/Node.js services that process X12 transactions (837, 835, 270/271, 999) including validation, coverage checks, payment status handling, and financial reconciliation. Built event-driven processing with Kafka/SQS and PostgreSQL using idempotent workflows, indexing, and consistent error handling. Created authenticated React/TypeScript interfaces for member-facing coverage and claims/payment views, and supported AWS deployments with Docker, CI/CD, and automated tests for coverage rules, claims transforms, and billing calculations.