Senior AI Full-Stack Engineer

Built voice AI agents for healthcare and SaaS clients using LiveKit, Pipecat, Deepgram, and ElevenLabs, reducing response time by 30–40% and achieving a 4.8/5 satisfaction score. Designed multi-stage RAG systems for clinical, insurance, and knowledge-retrieval workflows using LangChain, LangGraph, Pinecone, and Qdrant.

Developed multi-agent automation systems with shared Redis memory, tools, and guardrails for healthcare intake, insurance pre-authorization, CRM/EMS entry, scheduling, and lead workflows. These systems reduced manual processing by about 50% and increased qualified lead volume by 30–40%.

Fine-tuned open-source LLMs with LoRA/QLoRA, managed reproducible experimentation through MLflow and Weights & Biases, and delivered enterprise LLM solutions using AWS Bedrock, Azure AI Foundry, Azure OpenAI, Azure AI Search, and LangSmith.

Built FastAPI, Spring Boot, ASP.NET Core, Rust, and Ruby on Rails services for enterprise integrations, automation pipelines, low-latency stream processing, and rapid MVP delivery. Delivered clinical LLM and EHR integrations for Glass Health and multi-agent transcription and CRM workflows for Ringfree.

Led engineering for a seven-person team building an AI social network on AWS Bedrock and LangGraph for millions of daily users across web, iOS, and React Native.

Architected autonomous agent pipelines supporting thousands of concurrent AI characters, including multi-turn memory, persona consistency, safety filtering, prompt orchestration, and LLM-as-judge evaluation. Implemented generative character creation using ComfyUI, Stable Diffusion, and custom LoRA training pipelines.

Designed event-driven microservices for image generation, feed updates, and notifications, sustaining sub-second feed latency at millions of AI-generated events per day. Built Node.js and Supabase services with row-level security, authentication, and rate limiting, and contributed Java ingestion services and C#/.NET analytics tooling.

Led a four-person platform team building HIPAA- and SOC2-aligned AI systems that processed thousands of patient encounters each month.

Built clinical document pipelines using LLM and vision APIs to ingest intake forms, insurance documents, and clinical notes into pgvector and Pinecone. Developed a real-time AI medical receptionist using Twilio, ASR/TTS, and LLMs for identity verification, triage, routing, and patient intent capture.

Developed pre-visit voice and form ingestion workflows that generated structured SOAP notes and reduced time per clinical encounter. Built FastAPI microservices with RBAC, PHI/PII isolation, environment segregation, and audit logging, alongside ASP.NET Core EHR integrations and internal administrative tools.

Built real-time IoT data pipelines for the Cyclops retail intelligence platform, ingesting RTSP camera streams, processing more than 150,000 daily videos, and combining POS and sensor data for foot-traffic heatmaps and peak-hour alerts.

Developed OpenCV-powered demographic tracking and anonymized video-stream processing with role-based dashboards for retail and mall operations. Created script-to-video pipelines using FFmpeg and text-to-speech to automate shopper-behavior reporting.

Implemented cloud-rendered BI reporting with AWS MediaConvert and Three.js, reducing analytics delivery time by approximately 60%. Built Ruby on Rails REST APIs and Sidekiq workers for multi-tenant data ingestion and alert delivery.