Skip to content

AI development & integration · New York

Put AI to work on the jobs your team repeats every day.

I build AI assistants and automations that plug into the tools you already use: your inbox, your docs, your CRM, your support desk. Small, scoped projects that ship in weeks and keep a human in the loop where it matters.

  • Claude & OpenAI APIs
  • Next.js
  • TypeScript
  • Supabase
  • Remote or NYC on-site
support-intake.runexample run
  1. 09:14:02gmailnew message · "Order #4417 arrived damaged"
  2. 09:14:03classifyintent=return · priority=high · conf 0.94
  3. 09:14:03lookupshopify order #4417 · delivered 2d ago
  4. 09:14:05policyreturns.md §2 · within 30-day window ✓
  5. 09:14:07draftreply + prepaid label link
  6. 09:14:07route→ #support · needs approval
  7. 09:16:41humanapproved by Dana · sent ✓
  8. 09:16:42logticket closed · 2m 40s end to end
8 steps · 1 human checkwas ~25 min by hand

What I build

Four kinds of work, all wired into your existing stack.

No new platform for your team to learn. The AI shows up where people already work and hands off to a person when it isn't sure.

  • 01 · Assistants

    AI assistants trained on your business

    A chat assistant for your site or your team that answers from your own docs, policies, and product data, with sources, and escalates what it can't answer.

    e.g. website support bot · internal policy Q&A in Slack

  • 02 · Automation

    Workflow automation

    Inbox triage, lead intake, document processing, follow-ups. The repetitive reading-and-routing work that eats hours every week, handled by an AI pipeline you can audit.

    e.g. email → classify → draft → approve → send

  • 03 · Integration

    AI inside the tools you already pay for

    Connect a model to Gmail, Google Workspace, Slack, HubSpot, Shopify, Notion, or your own database through clean APIs and MCP servers, with permissions scoped tightly.

    e.g. CRM notes written from call transcripts

  • 04 · Custom apps

    Internal tools with AI built in

    Dashboards, portals, and back-office apps where AI does the heavy lifting: summarizing, extracting, searching, and drafting. Built in Next.js, hosted on Vercel, yours to keep.

    e.g. quote builder · document search portal

How a project runs

Start small, prove it works, then expand.

Every engagement begins with one workflow and a number we both agree on, like hours saved per week or response time.

  1. 30 min · Free

    Step 1: Audit call

    We walk through where your team's time goes and pick the one workflow where AI pays off fastest.

  2. ~1 week

    Step 2: Prototype

    A working version on your real data, so you can judge the output before committing to a full build.

  3. 2–4 weeks

    Step 3: Build & integrate

    Production version connected to your tools, with logging, approval steps, and guardrails.

  4. Ongoing

    Step 4: Hand off & support

    Docs your team can follow, training, and a monthly support option to tune prompts and add workflows.

Selected work

AI products I've shipped.

Three apps with Claude built in. The code is on GitHub.

Desktop app · AI assistant

ExiledOrb (GitHub, opens in a new tab)

A Path of Exile 1 & 2 overlay with price checks, a map-run tracker, an OAuth character viewer, and a Claude-powered trade assistant. Approved by the game's developer and running on a live API key.

AI: answers trade questions from live market data

  • Rust
  • Tauri 2
  • React
  • SQLite
  • Claude API

Desktop app · AI advisor

Pocketwatcher (GitHub, opens in a new tab)

A local-first personal finance app covering accounts, bills, budgets, debts, and goals. A rule engine ranks what to fix first, and a Claude advisor answers questions using your own numbers. Your data never leaves your machine.

AI: advice grounded in a private data snapshot

  • Rust
  • Tauri 2
  • React 19
  • SQLite
  • Claude API

Desktop app · AI tutor

StudyGuide (GitHub, opens in a new tab)

A study app for IT certifications. Claude writes quiz questions from each lesson's objectives, spaced repetition schedules the reviews, and XP and streaks keep you coming back. Unit-tested core, with installers built by GitHub Actions.

AI: generates questions and explanations per lesson

  • Tauri 2
  • React 19
  • TypeScript
  • Claude API

More on GitHub ↗ (opens in a new tab)

About

I'm Sean Mangin. Before I built AI systems, I spent years on the other side of the ticket queue.

I've handled 150+ support interactions a day in IT and customer support, and spent five years in merchandising execution at Home Depot. That's where I learned what actually slows a team down: the same questions, the same lookups, the same copy-paste between systems.

Now I build the tools I wish those teams had. I'm self-taught across the full stack, I build with Claude every day, and I design every automation the way a support lead would want it: logged, reversible, and with a person signing off on anything customer-facing.

Models
Claude · OpenAI
Frontend
Next.js · React · Tailwind
Backend
TypeScript · Supabase
Desktop
Rust · Tauri
Hosting
Vercel
Based in
Staten Island, NY

Get in touch

Tell me the task your team hates most.

Send a short note and I'll reply within one business day with whether AI is a good fit, and what a first prototype would look like.

Helpful to include

  • The workflow in a sentence or two
  • Roughly how many times a week it happens
  • The tools involved (Gmail, Slack, your CRM…)
  • What a good result would look like