Carlo Pelosi — Full-Stack Software Engineer · Phoenix, AZ

Full-stack engineer who builds the whole system.

Five years of full-stack experience — four of them at Starbucks building production applications and the Azure infrastructure underneath, in React, Node, and MongoDB. Then I kept going: a fleet of live client sites, a field-service platform a real company runs on, and a multi-agent AI orchestrator I built from scratch. I build things that have to work Monday morning, and then I keep them working.

View the work Download résumé

4 yrsFortune-100 production engineering

24/7AI systems running in production

180+service accounts run on my software

2024Starbucks Bravo Award

Selected work

Systems I've built and still stand behind.

Not demos. Every one of these runs — or ran — in production, with real users and real consequences when it breaks.

Case study — 01

Fortune-100 production infrastructure

Professional · 4 years

Enterprise Web Platform on Azure

Full-stack platform work at Starbucks: React and Node on a segmented Azure VM fleet, SAML SSO against an enterprise IdP, a Cosmos-to-Mongo migration, and NiFi pipelines feeding ERP systems.

Backend API never publicly exposed — private NSG rules, reachable only from the frontend subnet inside the VNET
SAML SSO federated against Starbucks' Microsoft IdP, with an Express BFF keeping tokens off the client
Migrated the backend from Cosmos DB + Redis to MongoDB, simplifying the stack and cutting cost

React
Node.js / Express
MongoDB
Azure VMs + App Service
Azure Front Door
SAML / Passport.js

Read the case study

Case study — 02

Field-service software for a company I co-own

Founder / Operator · Ongoing

McCool's Operations Platform

Internal operations platform for an Arizona pool-service company with 180+ accounts: a mobile-first technician portal, admin review workflows, performance dashboards, and an AI lead-intake bot — all built, shipped, and maintained by me.

Runs a real business: 180+ accounts and a field team depend on it weekly
Mobile-first PWA technician portal with role-based views and job submissions
Admin review workflows replacing texts, memory, and scattered spreadsheets

Next.js
React
TypeScript
MongoDB
PWA
Chart.js

Read the case study

Case study — 03

A custom multi-agent orchestrator, built from scratch

Self-directed · 2026 — active

AFO Agent OS

A 24/7 multi-agent AI system running my agency's operations: a Node orchestrator and workers on a VPS, PocketBase for state and memory, a Next.js command center on Vercel, and Telegram as the interaction surface — with an approval gate enforced in code.

Chose to build a custom orchestrator instead of adopting an off-the-shelf agent runtime — and can defend why
Consolidates all agent output into a single digest with a human approval gate enforced in code, not convention
Deliberate placement of every piece: agent loop on the VPS, UI on Vercel, no open inbound ports on the VPS

Node.js
Next.js
PocketBase
Telegram Bot API
VPS (Linux)
Claude / LLM APIs

Read the case study

See all work

How I work

The patterns behind the projects.

I build to own it.

Given the choice between renting a black box and building the thing myself, I build. Not for pride — because when it's 2 a.m. and something's down, I want to be the person who understands every layer.

I think in systems and data flows.

Where does each piece live, what talks to what, what's exposed, what happens when it fails. I reach for an architecture diagram before I reach for a framework.

Security is a default, not a feature.

Network segmentation, non-root service users, locked-down secrets, closed ports. It shows up in my work unprompted because I've been the one responsible when it mattered.

I ship pragmatically.

Static generation when static is enough. Managed services until self-hosting earns its keep. Tight v1s, everything else specified but dormant. Speed of shipping is a design constraint, not an excuse.

Technical range

What I work in — honestly tiered.

I'd rather you know exactly where I'm deep and where I'm still ramping. It makes for better interviews and better teams.

Tier 1 — Deep, production-proven

JavaScript / TypeScript
React
Node.js / Express
Next.js
Astro
Eleventy
HTML/CSS
MongoDB
Azure (App Service, VMs, Load Balancer, Front Door/CDN, VNET/NSGs)
PM2
CI/CD (Azure DevOps, GitHub)
Datadog / Splunk
REST API design
SAML SSO (Passport.js)
Box / Twilio / Mailgun integrations
Apache NiFi data pipelines
app security (SQL-injection hardening, network segmentation, cloud security reviews)

Tier 2 — Working, used in real projects

Microsoft SQL Server (schema design, stored procedures, query tuning)
PostgreSQL / MySQL
Python (agents & automation)
FastAPI
PocketBase / SQLite
React Native
Vue
PWA development
multi-agent orchestration
MCP servers
Claude Code workflows
GoHighLevel platform engineering

Tier 3 — Ramping, honestly

Java (strong OOP foundation, comfortable ramping)
C# / .NET Core / ASP.NET (actively learning)
Go (haven't shipped it in production yet — but I pick up languages fast, and I'd rather tell you that than pad the list)

Starbucks

Application Developer I → II · 4 years of full-stack and Azure infrastructure work.

Full experience

Full-stack engineer who builds the whole system.

Enterprise Web Platform on Azure

McCool's Operations Platform

AFO Agent OS

I build to own it.

I think in systems and data flows.

Security is a default, not a feature.

I ship pragmatically.

Tier 1 — Deep, production-proven

Tier 2 — Working, used in real projects

Tier 3 — Ramping, honestly

Let's talk about what you're building.