Marco
Nahmias.
Solo founder and full-stack engineer. I ship production web + AI systems end-to-end for real paying clients, using Claude Code as my primary development methodology. Previous exit (Medical Mall, 2017). Currently operating a multi-tenant SaaS platform serving hospitality, real estate, and construction clients — architecting, building, integrating, and supporting live systems alone, at the velocity of a small team.
A v0-style builder for verticalized businesses. Pick a template, get a full Next.js 16 site plus the PMS / CRM / lead dashboard the business needs — in one pass. Real code, multi-tenant, AI-generated content, Studio inline editor.
Direct booking engine, PMS (front desk, housekeeping, rates, night audit), Channex channel manager for OTAs, Twilio-powered guest messaging, Tilopay for LATAM payments. Running boutique hotels in Costa Rica and the US.
Solo-built AI-native website + operations platform. Production live, real paying clients, real support.
- Architected the unified backend serving hotel, real estate, and construction verticals from one codebase. Tenant isolation, auth, and deployment designed so one operator supports many clients without linear time cost.
- Shipped Hotel Native — the hospitality flagship. Direct booking engine, PMS (front desk, housekeeping, rates), Channex channel manager for OTAs, Twilio messaging (Voice/SMS/WhatsApp), Stripe payments.
- Built a v0-style onboarding wizard: pick a template, drop brand info, get a full production Next.js 16 site + vertical dashboard in ~30 seconds. 14-step provisioner writes content with Claude and Gemini, uploads media to R2, seeds rate plans and admin users.
- Studio — inline visual editor written as a standalone workspace package. Click-to-edit text, swap images, rewrite sections with AI. Every edit audited, multi-tenant scoped, pushed to production in under a second.
- Operate as solo engineer, solo support, solo account manager across 13 live brands on 49 domains. Claude Code is the methodology, not an autocomplete.
Agent architecture and LLM evaluation for enterprise workflows.
- Evaluated LangChain, AutoGPT, and BabyAGI before building a custom agent architecture from scratch.
- Ran A/B evals across OpenAI, Anthropic, and Cohere — systematic prompt design, tool invocation reliability, output validation.
- Implemented Pinecone and Weaviate vector search — foundation for the pgvector RAG that later shipped in Hotel Native.
- Researched agent observability, safety patterns, and failure-aware design for production deployment.
High-traffic web at scale. React / Node / PostgreSQL.
- Led full-stack development of high-traffic applications.
- Architected SEO infrastructure and automation systems. Managed AWS with a focus on scalability and cost.
- Mentored junior developers, owned technical architecture decisions end-to-end.
HIPAA-compliant e-commerce for medical supplies. Acquired 2017.
- Built payment processing, inventory, automated fulfillment, multi-marketplace syndication.
- Scaled to successful acquisition. First production experience in a regulated industry.
I don't paste prompts at ChatGPT. I architect inside Claude Code — spec, scaffold, iterate, ship. The AI writes heavy implementation. I spend my time on the hard calls: architecture, tradeoffs, integration edges, UX.
I'd rather put a rough version in front of a real user today than ship something polished in three weeks that the user didn't actually want.
I make architectural decisions with agency. Some right, some wrong, all mine. Ticket-takers don't compound — owners do.
Hard-earned lesson. I scope, contract, and set expectations before writing a line of code.
Shipped, paid, in production.
I'm not a ticket-taker. I own outcomes. I make opinionated architectural decisions, move fast, cut scope when it earns its keep, and ship.
I've made payroll, carried clients through outages at 2am, negotiated contracts in two languages, and owned the buck. I know what being the last line of defense feels like — and I prefer to do it inside a team of people who think the same way.