Technical project manager with a track record of shipping automation infrastructure from zero — not prototypes, not demos. Real systems handling real bookings, real money, and real customers — deployed across multiple businesses.
Not side projects. Not demos. Production systems handling real bookings, real money, and real customers — every day.
Designed and deployed a full AI automation stack on a self-managed VPS — n8n pipelines, ChatGPT, Gemini, Claude AI integrations, Docker containers, real-time API connections to Meta, Google, and Plaid. Every workflow solves a real operational problem that used to require manual work.
Full ownership and operations of three restaurant concepts in Atlanta. The restaurants aren't a side note — they're where the infrastructure gets stress-tested against real customers, real revenue, and real stakes. Every system that runs here had to actually work.
End-to-end automation for multiple restaurants: large-party booking auto-responders, SEO-optimized Google review replies, weekly finance reports via Plaid, and Meta Ads optimization with Claude AI — all running in production across two restaurant concepts.
A 24/7 voice AI agent built on Retell AI and Claude that handles every inbound call — answers menu and FAQ questions from a live knowledge base, routes reservations by party size and timing, and transfers to a human when needed. n8n polls every 2 minutes, extracts details from transcripts, and fires WhatsApp and email alerts to the team within 2 minutes of every call ending.
A multi-company Laravel back office for inventory tracking and staff tip-out calculations — role-based access, per-company gratuity split rules, and session-scoped company switching for operators running more than one concept.
A self-hosted multi-account Meta Ads dashboard with live data, drill-down from campaign to ad level, pixel event tracking, and a full history of Claude AI optimization decisions — built on Node/Express with PM2.
Weekly Meta Ads review across multiple restaurant accounts meant manually cross-referencing pixel data, ad status, and budgets — slow, and easy to defer. Letting an AI change live ad spend unsupervised was not an option; every change needed a human decision before it touched real money.
Runs in production every Monday across both accounts, with budget changes (±20%) applied only within approved selections. A separate, independent daily monitor watches for spend spikes (≥1.5x trailing average) as a same-day safety net between the weekly cycles — informational only, no approval loop of its own.
During dinner service every host is on the floor. The phone goes unanswered, callers don't leave voicemails, and reservations walk out the door. A dedicated phone host for every shift isn't economical — and traditional IVR just frustrates callers.
Calls that used to fall through during dinner service now reach Maya regardless of what the floor is doing. Reservation details — name, party, date, time, email, occasion — are extracted and delivered to the team via WhatsApp within 2 minutes of every call ending. Confirmation calls and complex requests route immediately to a human with no friction.
Defined end-to-end product vision and 3-phase technical roadmaps from concept through production — including full specs covering data models, user flows, and phased delivery plans for both internal systems and external products.
Led architecture decisions for mobile-first web applications, dual-purpose API endpoints, multi-party ledger systems, and role-aware routing across multi-tenant platforms — making the right call at the right time without over-engineering.
Managed cross-functional build scope across frontend, backend, and payment infrastructure workstreams — sequencing delivery to de-risk complexity and defining acceptance criteria, non-goals, and phase boundaries to keep velocity up.
Designed and deployed Claude AI integrations in production across voice agents, booking automation, ad optimization, review response, and financial analysis — including failsafe patterns (continueOnFail, transcript fallbacks, timestamp deduplication, daily publish cycles) that keep systems stable under real load.
Built and shipped a production voice AI host (Retell AI + Claude) handling inbound restaurant calls 24/7 — prompt engineering for reservation routing logic, knowledge base integration for live menu retrieval, daily automated prompt updates via n8n, and a real-time notification pipeline delivering call summaries to WhatsApp within 2 minutes of every call.
Open to conversations about technical project management roles, founder partnerships, or helping other hospitality operators build AI infrastructure for their businesses.