10+ years shipping production LLM, RAG, and analytics systems on AWS and Azure — including inside legacy infrastructure that couldn't go down. I build AI that moves metrics, then lead the teams that maintain it.
Built from scratch using n8n, GPT-4o, and Postgres-backed conversation memory. Qualifies hiring and project intent, captures contact info, scores leads, logs full transcripts, and fires an SMS notification on qualified conversations — running live on this page right now.
Self-hosted support chatbot with a .NET adapter translating natural language into SQL queries against a legacy CRM. Deployed across live chat, email, SMS, and voice — no ERP re-platform required.
Conversational AI workflow with persistent session memory stored in Postgres. Full transcript logging enables quality review, audit trails, and retraining data capture from real customer interactions.
PyTorch + ViT-based classification pipeline trained on 40K+ matched design files to automate embroidery digitizing decisions. Includes placement-aware size priors and a Claude API suitability agent for pre-screening customer artwork.
Source is private — request a walkthrough ↗
Production MCP server exposing internal SQL Server and Postgres databases to AI tools over HTTPS. Consumed by Claude Desktop, n8n workflows, and a custom React backend. Secured with Cloudflare Access + Azure AD authentication.
Source is private — request a walkthrough ↗
4-workflow n8n pipeline that takes a product image and publishes live Meta ads — no human steps. GPT-4o-mini analyzes images, GPT-4.1-mini writes prompts, gpt-image-1 generates 4 ad creatives in parallel. Meta Graph API v22 publishes with dynamic creative optimization.
Source is private — request a walkthrough ↗

Here's something I say in almost every AI conversation that makes people uncomfortable: AI isn't as reliable as you think it is.
It hallucinates. It drifts. It performs beautifully in a controlled environment and does something completely unexpected the moment a real user touches it. The companies winning with AI right now aren't the ones who deployed a model and walked away — they're the ones who treat it like a production system that needs constant monitoring, retraining, and someone paying close attention. That's the lens I bring to everything I build.
Stitch America is a family business. I joined to modernize it — not "add a dashboard," but build the AI infrastructure from scratch, on top of a decade of legacy systems, with no greenfield luxury. When I built the RAG chatbot, I was connecting a language model to a legacy SQL CRM that wasn't designed to be queried in natural language, in a live production environment that couldn't go down. The test environment worked fine. The live deployment didn't. That gap is where I learned the most.
I've taken that work beyond Stitch America too — deploying a Claude-based multi-agent system for a pharmaceutical client's API modernization and security assessment, building AI-assisted websites for law firms, and handling digital automation for political campaigns at the state and federal level.
I stayed because the work was real and the stakes were real. I've led teams of up to 10, managed budgets over $2M, and still write the architecture. Now I'm ready to take what I built and lead a team doing it somewhere new.
Before Stitch America, I was on the Apple Maps global launch team in Knoxville — triaging geospatial defects at scale and onboarding technicians. The team hit the district's highest NPS (93) two quarters running. That's where I learned what shipping something real, to real users, actually feels like.
B.S. in Business Administration, University of Tennessee. IBM Machine Learning certification (June 2025). Currently completing AWS Solutions Architect – Professional and Azure AI Engineer.
I'm looking for my next role — AI Engineering Manager, AI Architect, or a first AI hire at a company ready to build seriously. Remote or hybrid in Atlanta. I respond to every message personally.