AI Full Stack Software Engineer
From event streams to user-facing decisions, I engineer the whole path.
400K+
50K+
99.99%
4+
Three layers I keep joined: services under load, AI that survives review, and interfaces people actually use.
Services under pressure
Event-driven services, API platforms, and cloud delivery—the paths data takes when traffic, retries, and failure are real.
Services that keep moving when traffic, data, and failure get messy.
Useful after the demo
LangGraph workflows, grounded RAG, and review loops—AI wired into product behavior with evaluation and guardrails.
AI that earns its place in the product—not just the demo.
Contract to screen
API contracts, data pipelines, and React interfaces as one ownership loop—not a handoff between layers.
From API contract to the screen someone actually uses.
I like hard boundaries, clear contracts, and systems that fail in ways you can diagnose.
From Kafka pipelines to the interface—someone has to own the whole route. I prefer that someone to be me.
I care about what happens after the demo: evaluation, feedback, and the boring controls that keep AI honest.
Clinician-facing AI copilot platform at Abridge — FastAPI backends, React/TypeScript review interfaces, LangChain RAG, and FHIR/HL7 clinical integrations.
01 — Problem
Clinical teams needed a unified platform to surface AI-generated insights from fragmented EHR data — with citation review, compliance controls, and production-grade reliability.
02 — Architecture
Microservices architecture with FHIR/HL7 ingestion → Kafka event pipelines → LangGraph RAG workflows → React clinician review interfaces, deployed on AWS EKS with Terraform and full observability.
03 — Intelligence
LangGraph workflows orchestrate multi-model RAG with citation review, groundedness checks, and human-in-the-loop approval before any insight reaches clinical workflows.
AI Summary
Patient history consolidated from 5 clinical domains. Recommended follow-up based on risk model output.
04 — Impact
400K+
Records
50K+
Interactions/mo
99.99%
Uptime
25%
Cost saved
Additional systems across AI workflows, full-stack apps, and data platforms.
Featured · Nov 2025
Full-stack AI dashboard analyzing 451 call center interactions with Gemini 2.5 — LLM insights, funnels, and revenue modeling.
Problem
Call center managers lacked actionable insight from hundreds of daily conversations.
Architecture
Express API + Gemini 2.5 extraction pipeline feeding React dashboards with funnel and revenue models.
Impact
Lab · Oct 2025
Custom RAG chatbot on AWS Bedrock and AgentCore for domain-specific Q&A over PDFs and text files.
Challenge
Domain-specific document Q&A required a managed retrieval pipeline without building from scratch.
What shipped
Lab · Jul 2025
AI concierge with RAG pipeline on Pinecone + Supabase, automating guest workflows via REST APIs.
Approach
RAG on Pinecone + Supabase with n8n workflow automation and REST API integrations.
Impact
Lab · Jan 2025
Real-time gesture recognition detecting 36 hand signs at 90%+ accuracy with MediaPipe and TensorFlow.
Problem
Real-time gesture classification needed a modular, retrainable computer vision pipeline.
Architecture
MediaPipe landmark extraction → TensorFlow classifier with modular training pipeline.
Impact
Lab · Mar 2024
Python + SQL CLI analytics tool for traffic incident, vehicle, road, and signal datasets.
Challenge
Multi-dataset traffic analysis required normalized schemas and repeatable query workflows.
What shipped
Swipe to explore · 1 / 5
From enterprise Java and cloud data platforms to production AI — owning more of the path each step.
Select a cluster to see the technologies behind each domain.
AI & Agents
12 technologies
Production technologies & capabilities
High-signal stack
AI, backend, data, cloud, and interface—designed as one working system.
Most interesting problems sit between layers—where an API boundary, a data path, and a UI decision have to agree. That's where I like to work.
Based in the USA. Curious by default, careful at the edges, and stubborn about software that still makes sense after launch.
Open to AI Full Stack, applied AI, and platform engineering opportunities.
Usually responds within 24 hours.
Have a role, idea, or project in mind? Send me a message.