Open to IT, AI and cybersecurity roles

I build with AI and keep systems secure.

I'm Chance Hart. I use Claude Code every day to build automations, connect tools through APIs and ship working software. Before that I helped customers in retail and by phone, ran office admin and handled sensitive client data. I know how to make a careful fix and how to explain it.

chance@portfolio: ~● live

$ whoami

Chance Hart. AI automation, IT support, security-minded.

$ cat now.txt

Building tools with Claude Code and studying for CompTIA Security+.

$ ls projects/

home-soc-lab/ ai-second-brain/ claude-session-saver/ store-api/ this-site/

# Type a command below, or tap one. Try: help

Looking for
IT support, AI automation, junior security analyst
Strongest proof
Home SOC, always on in the cloud: quarantines malware in ~4 s, blocks bad IPs in ~6 s
Based in
Alabama, US · Central Time
Work style
Remote, hybrid or on-site

Selected work

Case studies

Each one is a real system I built and use. Claude Code is my pair programmer: I set the goal, read the code it writes, and test until it works.

Case 01 · Security operations

A home SOC that catches, blocks and cleans up threats

My own small security operations center, running 24/7 on Oracle Cloud ($0) and reachable only through my private Tailscale network. It watches my PC, decides what is malicious and responds on its own, like a lightweight McAfee that I built, tested and can explain line by line.

Wazuh SIEMSysmonYARAVirusTotalMITRE ATT&CKDockerOracle CloudTailscalePowerShell
~4 sto quarantine a bad file
~6 sto block a malicious IP
17,000+known-bad IPs tracked daily
300+alerts triaged, 0 attacks
0ports open to the internet

Problem

Security teams want proof that you can run a SIEM, not just name one. I wanted a real system I could attack safely, watch and tune, the way a SOC analyst does every day.

What it does

  • Locks away bad files in about 4 seconds, and can restore them (YARA rules + VirusTotal's 66 antivirus engines)
  • Blocks bad IPs in Windows Firewall in about 6 seconds, using threat-intel feeds refreshed every morning
  • Warns about ransomware: ransom notes, encrypted files, sudden bursts of file changes
  • Runs a full scan every week and shows Windows pop-ups when it acts
  • Doubles as my personal VPN: one tap on my phone sends all traffic through the server, encrypted with WireGuard (Tailscale exit node)

Result

  • Every detection is mapped to MITRE ATT&CK and proven by a safe test script
  • Triaged 300+ alerts on day one: no attacks, and 223 false alarms cut with narrow tuning rules
  • Replaced the vendor's default admin and API passwords, then confirmed the old ones were rejected
  • Pen-tested it myself: a full 65,535-port scan finds nothing open to the internet; fixed every finding (SSH locked to Tailscale, containers blocked from the cloud metadata service, secrets root-only)
terminal · quarantine testView on GitHub ↗
$ make-test-file.ps1
Created downloads\homelab-yara-test.txt
# about 4 seconds later
homelab-av: QUARANTINED reason="YARA: HomeLab_Test_Marker"
Alert (level 12): malicious file quarantined   T1204.002
$ restore-quarantine.ps1 -Id 076eebb52a50
Restored downloads\homelab-yara-test.txt
Case 02 · AI agent workflow

An AI agent that remembers every session

My notes system in Obsidian, run by Claude Code. Every work session starts where the last one stopped, on any device.

PythonClaude Code hooksUnit testsOpen source

Problem

AI chats forget everything when they end. I was losing decisions and repeating context every time I started a session.

What I built

  • A Python hook that runs after every AI reply and saves the session as a searchable note
  • A handoff file the agent updates after every decision
  • Weekly and monthly reviews that run on a schedule

Decisions

  • Privacy: the hook saves the conversation only. Tool output and system messages are left out.
  • Safe writes: it writes to a temp file and then swaps it in, so cloud sync never picks up a half-written note.
  • Open source: I published it for anyone to use. See Case 03.
Claude Code sessionPC or phone
Stop hooksave_session.py
Markdown transcriptJournal/Sessions/
OneDrive vaultsynced to every device

Plus: scheduled AI tasks → weekly and monthly reviews written into the same vault.

.claude/hooks/save_session.pyView full code on GitHub ↗
def main():
    payload = json.load(sys.stdin)
    path = payload.get("transcript_path")
    ...
            if e.get("type") not in ("user", "assistant") or e.get("isMeta") or e.get("isSidechain"):
                continue
            ...
            text = STRIP.sub("", text).strip()
    ...
    tmp = os.path.join(OUT_DIR, name + ".tmp")
    with open(tmp, "w", encoding="utf-8") as f:
        f.write("\n".join(out))
    os.replace(tmp, os.path.join(OUT_DIR, name))
Case 03 · Open-source tool

claude-session-saver: shipping it for everyone

I turned the hook from Case 01 into a public tool anyone can install with one command, with tests, automated CI on three operating systems, and versioned releases.

Python CLIunittestGitHub Actionspip / pipxMIT
v1.1.13 releases
25unit tests
9 / 9CI jobs passing
3OSes: Windows, macOS, Linux
0dependencies

Problem

My hook only worked in my own vault. To be useful to others it needed a real installer, safe defaults, and proof that it works on any machine.

What I built

  • A CLI: install, uninstall, backfill (import past sessions)
  • An installer that merges into existing settings, keeps a backup and never duplicates itself
  • Secret redaction: API keys, tokens, private keys and passwords become [REDACTED] before saving

Result

  • Installs from GitHub in one command; tested in a clean environment
  • Found and fixed a real bug: installed paths broke on synced folders, so I shipped a --portable mode in v1.1.0
  • Found and fixed a second bug from real use: notes scattered into subfolders when the agent changed folders. Fixed in v1.1.1 with a regression test
  • Docs: README, changelog and a security policy
terminal · installView on GitHub ↗
$ pipx install git+https://github.com/ChanceHart/claude-session-saver
$ claude-session-saver install --portable
Installed the hook in .claude/settings.json
Backup of the previous settings: settings.json.bak-20261005225402
$ claude-session-saver --version
claude-session-saver 1.1.1
Case 04 · API integration + web

Running a store through an AI agent and an API

My e-commerce brand, Ordelle. I connected Claude to Shopify, so the agent does the admin work through the API instead of clicking through screens.

MCPGraphQLPythonLiquid / CSS / JS

Problem

Setting up a store by hand is slow and easy to get wrong: dozens of products, collections, tags and theme edits.

What I built

  • Products and collections created through the Shopify Admin GraphQL API, using MCP
  • Python scripts that build and update the theme files
  • Store features written in Liquid, CSS and JavaScript

Result

  • 23 products and 9 auto-updating collections created through the API
  • Tested on desktop and phone, with no sideways scrolling
  • A product safety review applied to the whole catalog
Claude Codeagent
MCP connectortool calls
Shopify Admin APIGraphQL
Storeproducts, collections, theme

Theme code: build_theme_files.py → Liquid / CSS / JS → draft theme, tested before publishing.

Case 05 · This site

The page you're reading

Built with Claude Code: hand-written HTML, CSS and JavaScript, with no framework and no tracking. The design follows Twingate's published design system.

HTMLCSSJavaScriptAccessibility

Built

A working terminal, a layout that works on phones, and design tokens taken from a real security company's system.

Accessible

Keyboard friendly, with visible focus. Animation turns off if your device asks for reduced motion.

Private

The page sets no cookies and loads no trackers or analytics. My phone number and address are left off on purpose. More below.

How I work with AI

Three rules I follow

01

Check everything

AI writes fast and sometimes writes wrong. I read the output, test it and only trust what works.

02

Share less data

I give tools only the data they need and keep personal information out of logs and public pages.

03

Automate the repeats

If I do something by hand twice, I write a script or schedule a task for it the third time.

Skills

Hands-on with AI, solid on the basics

I learn by building. I pick a real problem, build a fix with AI, then test it until it works. I also know how to support the people who use the systems.

AI and automation

I use AI agents to build real tools.

  • Claude Code and AI agents
  • MCP tool connections
  • Hooks and scheduled tasks
  • Prompt writing and review

Building and code

I read code and make it run.

  • Python scripts and CLIs
  • Unit tests + GitHub Actions CI
  • HTML, CSS, JavaScript
  • GraphQL APIs

IT support

I fix problems and explain them clearly.

  • Troubleshooting
  • Basic IT support
  • Office software and data systems
  • Helping users by phone and in person

Security operations

I run my own home SOC, and I've been trusted with sensitive data.

  • Wazuh SIEM, Sysmon and alert triage
  • YARA, VirusTotal and threat-intel feeds
  • Detections mapped to MITRE ATT&CK
  • Client health and financial records handled safely
Wazuh SIEMSysmonYARAMITRE ATT&CKDockerOracle CloudTailscaleClaude CodeClaudeMCPPythonHTML / CSS / JSGraphQLGit / GitHubGitHub ActionsunittestShopify LiquidJSONWindowsObsidianCRM systemsMicrosoft Office

Experience

Where I learned to fix things for people

  1. Sep 2026 – Present

    Founder

    Ordelle · Self-employed · Remote

    • Founded Ordelle, a home-organization brand built with AI: brand identity and logo, a Shopify store with 23 product listings, store policies, a financial model and a Google/Meta ad plan.

    RelevantOwnership, planning, running systems end to end.

  2. Sep 2026 – Present

    AI Automation Developer (Independent)

    Self-employed · Remote

    • Built a home security operations center (home-soc-lab): Wazuh SIEM running 24/7 on Oracle Cloud behind Tailscale, with Sysmon on my PC; quarantines bad files in about 4 seconds (YARA + VirusTotal), blocks bad IPs in about 6 seconds from daily threat intel and warns about ransomware.
    • Triaged 300+ alerts (no attacks; 223 false alarms tuned out), mapped every detection to MITRE ATT&CK with a safe test, and pen-tested the server: zero ports open to the internet.
    • Built and open-sourced claude-session-saver: a Python CLI that gives Claude Code a memory, with secret redaction, 25 unit tests and GitHub Actions CI on Windows, macOS and Linux.
    • Connected Claude to the Shopify Admin GraphQL API through MCP; created 23 products and 9 auto-updating collections and wrote Python scripts that build the store theme.
    • Built and shipped this portfolio on GitHub Pages: hand-coded HTML, CSS and JavaScript with a strict Content Security Policy and no trackers.

    RelevantAI automation, APIs, testing, shipping software.

  3. Apr 2026 – Sep 2026

    Retail Associate / Marketing Specialist

    Jubilee Pool & Patio · Daphne, AL

    • Worked on an AI-agent dashboard to track KPIs and automate reports for management.
    • Kept customer records accurate in the CRM; managed inventory and stock orders.
    • Ran e-marketing: promotions, social media and the store's online presence.
    • Processed payments and bank deposits, and solved customer problems in person and by phone.

    RelevantData accuracy, business systems, user support.

  4. Jul 2025 – Oct 2025

    Life and Health Insurance Agent

    Globe Life Liberty National · Mobile, AL

    • Handled clients' sensitive health and financial information and kept records organized.
    • Explained complex coverage options in plain language in one-on-one consultations.
    • Managed my own pipeline, scheduling and follow-ups.

    RelevantHandling sensitive data, clear explanations.

  5. Dec 2024 – Jul 2025

    Administrative Assistant

    Long Leaf Services LLC · Daphne, AL

    • Answered customer calls, scheduled jobs and followed up with customers.
    • Created invoices, tracked billing and kept job records organized.
    • Coordinated job details between customers and field crews.

    RelevantCustomer communication, scheduling, accurate records.

Training

Learning AI and security in public

  1. Oct 2026

    Generative AI Fundamentals

    Databricks · completed October 2026

    • Covered how large language models work, generative AI use cases, and the risks and responsible use of AI in a business.

    RelevantBacks up the AI automation work on this site with a recognized AI credential.

  2. Oct 2026 – Jan 2027

    CompTIA Security+ (SY0-701)

    In progress · exam planned for January 2027

    • Studying the official objectives: threats and attacks, security architecture, security operations, and governance and risk.
    • Practicing hands-on with a home security lab, using a free security monitoring tool (Wazuh) to collect and analyze logs.

    RelevantThe standard entry-level security certification for SOC and analyst roles.

Security notes

How this site handles your data

I want to work in security, so I built this site the way I'd want any site built.

No trackingThis page's code sets no cookies and loads no analytics or ad pixels.
No third-party scriptsAll JavaScript is written inline on this page. The only outside request is for fonts.
Less personal dataOnly my email is shown, as plain text. No phone number or home address.
Nothing you type is sentThe terminal runs only in your browser. This page's code doesn't send or save what you type.

Contact

Hiring for IT, AI or security?

Email me and I'll get back to you quickly.

Email
ichart0421@gmail.com
LinkedIn ↗ Resume ↗