2017 — 2024
Orange Coast College
Computer Science coursework
4.00 GPA
President's List & Honors List. Coursework: C++ I & II, Java, C#/.NET, Python, JavaScript, and data structures.
~$ whoami · AI automation engineer
I build AI systems that do the repetitive work of running websites.
I use them to run 18 live sites myself: 13 of my own and 5 for clients. They send each AI task to the cheapest model that can do it and stop at a spending limit, answer visitors’ questions on ten local directory sites, and report new pages to Google and Bing every day.
Building with: multi-model routing
Open to AI engineering roles · Remote (U.S.) · Available now
portfolio navigation (not a real shell)
# Portfolio navigation — not a real shell.
# Type a command or press Tab. Try `ls`, `help`, `hire`.
# Esc leaves the terminal. Shift+Tab goes back.
$
18
Live sites I run
13 my own · 5 built for clients · September 2026
10
Sites with an AI chat assistant
Each answers visitors’ questions about local services — live today
1,300+
Pages reported to search engines in one run
Accepted by IndexNow across 15 sites I run · scheduled job, Sept 17, 2026
Built with
- Next.js 16
- React 19
- TypeScript
- Tailwind 4
- Supabase
- OpenAI
- Vercel
in plain english
What I’ve built with AI
Four systems, explained without jargon. Each one runs today, names the tools it uses, and links to something you can check yourself.
Chat assistants on ten live websites
Ten Rochester directory sites each have an assistant that answers visitors’ questions about local services, day or night, and saves contact requests as leads.
Tools
- OpenAI GPT-4o mini
- Next.js
- MongoDB
Result
Live on all 10 sites — checked September 17, 2026.
A traffic controller for AI models
Every AI task goes through one program I wrote. It picks the cheapest model that can do the job — a free one on my own PC for routine work, Claude or GPT for hard judgment — switches to a backup if one fails, and stops when a spending limit is hit.
Tools
- Claude
- GPT
- Gemini
- Ollama (local models)
- LiteLLM
Result
A hard dollar cap on every run, enforced in code rather than by habit.
Search-engine updates that run themselves
A scheduled daily job tells Google and Bing about new and changed pages across the sites I run, so nobody has to do it by hand.
Tools
- IndexNow
- Google Indexing API
- Search Console API
- Node.js
Result
More than 1,300 pages across 15 sites accepted by IndexNow in one run on September 17, 2026.
Open-source code anyone can read
Public projects on GitHub. One lets a chatbot answer from your own documents and measures how often it finds the right one. Another checks that an AI agent really produced the files it says it did — because AI tools sometimes report work they never finished.
Tools
- TypeScript
- Ollama
- Vitest
- GitHub Actions
Result
On a public test of 300 science questions (BEIR SciFact), the right source lands in the top 10 results about 85% of the time — Recall@10 of 0.846, from the same run that lifted nDCG@10 from 0.687 to 0.712.
All of this was built at ROC Web Tech, the web studio I founded. rocwebtech.com
01 / work
Selected work
The AI platform that routes every model call behind the sites I run. A members-only lounge with sign-in, role-based access, and an admin panel. A garage door company site built for lead capture. All three are running in production.
2026 · Live
Multi-Model AI Orchestration Platform
The AI infrastructure behind the sites I run — multi-model routing, RAG, chat assistants, and budget-enforced automation
- TypeScript
- Node.js
- LiteLLM
- Ollama
2026 · Live
My Place Cigar Lounge
Private members-only cigar lounge with a custom marketing site, member portal, and admin dashboard
- Next.js 16
- TypeScript
- Tailwind CSS 4
- Supabase (auth + Postgres)
2026 · Live
Monroe Overhead Door
Lead-gen site for a 4.9-star Monroe County garage door company — 19 service-area pages, IndexNow auto-ping
- Next.js
- TypeScript
- Tailwind CSS
- Vercel
02 / story
How I build
This spring I left RIT to build full-time. Today I run 18 live sites — 13 of my own and 5 for clients — on an automation layer I wrote: a multi-model AI router with automatic failover, a RAG pipeline that lets chat assistants answer from each site’s own content, those assistants on ten directory sites, and a daily search-indexing job that runs itself.
I own the full loop and I work directly with the business owners I build for: scope the work, design the UI, write the code, integrate APIs and AI, deploy on Vercel, and keep supporting it. Behind it are CS fundamentals from Orange Coast College (Computer Science coursework, 4.00 GPA) — data structures, discrete math, and Python/Java/C++/C#. I measure before I claim: my public RAG pipeline ships with a benchmark on a standard retrieval dataset, and my Supabase auth reference proves its row-level-security policies with integration tests.
ROC Web Tech is the web studio I founded to do this work. When I join a team, I stop taking new client projects and keep supporting the clients I already have. Shipping, supporting and automating real sites for real owners is the experience I’m bringing to a team.
Working with
- Claude Code + MCPs
- n8n orchestration
- Multi-model agents
- RAG pipelines
03 / capabilities
Capabilities
Next.js + TypeScript core, AI-augmented workflow, automation-first delivery.
AI, Automation & Data
Production AI systems — multi-model routing with failover, RAG (retrieval that lets a chatbot answer from your own documents), chat assistants, and n8n workflow automation on real databases.
- Multi-model routing
- RAG + hybrid search
- n8n automation
- Claude / OpenAI / Ollama
- MCP tool servers
- Supabase / Postgres
Full-Stack Web
Production Next.js 16 apps with App Router, Server Components, strict TypeScript, and optimized bundles.
- Next.js 16
- React 19
- TypeScript
- Tailwind 4
- Node.js
- REST APIs
- Auth.js & RBAC
- Stripe
Practice & Delivery
Accessible and fast by default: Core Web Vitals, WCAG, structured data, and one-click Vercel deploys, with Vitest and Playwright tests where the logic lives.
- Core Web Vitals
- WCAG a11y
- SEO / JSON-LD
- Vitest & Playwright
- Git & GitHub
- Vercel CI/CD
Languages & CS
CS fundamentals across multiple languages — data structures, algorithms, discrete math, and systems-level coursework.
- TypeScript
- JavaScript
- Python
- Java
- C / C++
- C#
- Data Structures
~$ git log --author=connor
Building in the open
I build and ship constantly. Most production work lives in private repos. The public side is small on purpose: three projects that show how I build, test, and measure. An architecture walkthrough of the private work is available in interviews.
- rag-pipeline-ts Lets a chatbot answer from your own documents — with tests, CI, and a reproducible retrieval benchmark
- supabase-rls-auth-demo Sign-in, roles, and database access rules, with integration tests that prove the rules hold
- assert-artifacts Checks that the files an AI agent says it wrote really exist — zero dependencies, CI on Windows and Linux
04 / writing
Build logs & engineering notes
How I ship production sites fast — the stack, the process, and the tradeoffs, written so you can copy them.
Sep 17, 2026 · 7 min read
The CSS mask that cost my home page 400ms (and four wrong guesses first)
One route scored 91 while two others on the same site scored 98 and 97. Four plausible causes measured and disproved, the real one found by elimination — and the follow-up measurement that failed its own control.
- performance
- LCP
- CSS
Jul 16, 2026 · 5 min read
The hero fade that broke LCP (and how I actually diagnosed it)
A fade-in entrance kept the hero from being the early LCP candidate, so the metric latched onto a font-swap repaint at ~3.3s. The diagnosis, the fix, the rule it became — and the measurement that still doesn’t add up.
- performance
- LCP
- debugging
May 16, 2026 · 8 min read
How I ship a production site in 5 days
What actually makes a five-day build possible — the stack decisions, where AI helps, where I refuse to let it, and the checks a page has to pass before it goes live.
- process
- workflow
- AI-assisted
05 / background
Education
CS fundamentals — data structures, algorithms, discrete math, and multiple languages.
Santa Barbara City College
Computer Science coursework
4.00 GPA
Vice President, Computer Science Club · Founding President, Entrepreneurship Club.
2024 — 2026
Rochester Institute of Technology
Computing & Information Technologies
3.4 GPA
Left in Spring 2026 to ship production work full-time. Coursework: Cybersecurity Foundations, Software Development & Problem Solving, discrete math, statistics, and networking.
Need a site like these?
I build these for Rochester businesses too
The sites above are client work, not demos — hand-coded, fast, and built to rank locally. ROC Web Tech is where that happens, with pricing published up front and a free technical audit of whatever you have now.
06 / let’s talk
Want an AI engineer who already ships?
I’m looking for an AI automation or applied-AI engineering role — remote (U.S.), hybrid or on-site, open to relocation, full-time or contract. You get someone who has already shipped production auth, payments, multi-model AI routing, RAG, and automation pipelines end to end. Résumé and GitHub are one click away; I reply within one business day.