Hi, I'm Caleb 👋

Full-Stack Developer | Founder @ Arcanet Consulting | Product-Focused Builder

• 💻 I build software across frontend, backend, and systems-level projects
• 🏢 Founder of Arcanet Consulting — hands-on engineering help for startups and small teams, from architecture to shipping
• 🤖 I’m especially interested in autonomous/intelligent browser workflows
• 🎯 Current focus: shipping impactful projects and improving software craftsmanship

Languages

Frameworks & Runtime

Cloud, Deployment & Data

Tools

Live: caleby03.github.io/Homelab-Portfolio · Repo: Caleby03/Homelab-Portfolio

A self-hosted homelab run as infrastructure-as-config, built around a deliberate split of responsibilities: an always-on mini PC for stateful services, a Windows workstation used as burst GPU capacity, and a Raspberry Pi as an independent watchdog. Every node is assigned work it is actually suited to, every container is capacity-bounded, and the monitoring is designed around the assumption that any node — including the one doing the monitoring — will eventually fail.

• Containerized service stacks split by lifecycle and failure domain rather than by subject matter, so alerting can't be taken down by the same command that takes down the thing it watches
• GitOps deployment — stacks deploy straight from a git repo and reconcile on their own, so a commit is the deploy
• Zero-trust networking over a WireGuard mesh: no port forwarding, no public ingress, access authenticated per device rather than per network
• Found and closed a live security gap — an audit revealed published container ports reachable from the LAN, because Docker inserts its own iptables rules ahead of the chain the host firewall manages, so a deny incoming policy silently did not apply. Moved enforcement into the DOCKER-USER chain and verified it across reboot, daemon restart, and from an off-network host
• Metrics and dashboards with recording rules pre-computing dashboard values, plus Architecture Decision Records written at the time each choice was made rather than reconstructed afterward

Stack: Docker · Docker Compose · Portainer · Prometheus · Grafana · Tailscale · iptables/UFW · PostgreSQL · Redis · Ubuntu

Site: arcanettechnology.com

My software engineering & technology consulting practice. I work directly with founders and small teams — hands-on, no hand-offs — to design, build, and modernize software, from architecture and cloud setup through shipping the actual product. The pitch is how I work: direct, honest scoping, and shipping without cutting the corners that come back later.

• Custom software — web apps, APIs, and internal tools built to be maintainable
• Cloud & infrastructure — architecture, migrations, and CI/CD on AWS, GCP, Azure, and Cloudflare
• Technical strategy — system design, code and architecture reviews, and honest build-vs-buy advice
• Product delivery — scoping, prioritizing, and shipping from MVP to launch

Engagements: project-based or ongoing · work directly with the person doing the work

Repo: Caleby03/HarmonyHouseWebapp (private)

A direct-booking and calendar-sync platform for Harmony House NC, a short-term rental in the Blue Ridge Mountains of North Carolina catering to wedding parties and equestrian guests. Guests book directly on the property's own site instead of paying platform fees on Airbnb or Vrbo.

• Live availability calendar with direct date selection and instant booking
• Square Web Payments for checkout; Resend for confirmation and cancellation emails
• Automatic calendar sync — a GitHub Actions cron ingests Airbnb & Vrbo iCal feeds every 15 minutes so availability is always accurate
• Host portal (/portal): reservation list, platform-color-coded calendar, revenue & occupancy metrics, property editor, and date blocking

Stack: Next.js 16 · React 19 · Tailwind v4 · FastAPI · SQLModel · PostgreSQL · Square · Resend · Railway

Repo: Caleby03/Intelligent-Browser-Agents · UCF Senior Design (Team L07)

A platform that turns natural-language instructions into real actions performed on a live browser session. Users delegate repetitive digital chores — filling forms, navigating dashboards, downloading files — to an agent that reads the page, plans, and executes visually and safely, with the whole run streamed back to the UI in real time.

• LLM-driven agent runtime built on LangGraph — an orchestrator → executor → verifier → fallback/interaction loop
• Playwright browser automation with ARIA-oriented DOM extraction and a modular action layer (navigate, click, type, search, scroll, press_key, wait, extract_content)
• Human-in-the-loop (HITL): the agent pauses to ask for clarifications, approvals, or credentials mid-run
• Live browser frames, logs, and agent reasoning streamed to the frontend over WebSockets
• Full account system + encrypted storage of reusable credentials and profile data for automation contexts

Stack: React + Vite · FastAPI · LangGraph · Playwright · PostgreSQL · WebSockets · AWS (LightSail/EC2)

Demo Video: Watch on YouTube · Showcase: UCF Senior Design page

_{Generated twice daily by a GitHub Action · contribution counts include private work; languages reflect public repos}