Technology operations / agentic AI systems

I build AI systems people can follow.

I build AgentReef and Kynetik through Angel Software Solutions while supporting airline Tech Ops systems at Avelo. Aircraft records and Apple release validation taught me that good software needs clear ownership, testable releases, and an answer for what happened.

  • Building AgentReef and Kynetik
  • OpenClaw Community Staff
  • Houston / open to Dallas-Fort Worth
Current system map quiet mode
AgentReef

Routes work between agents and keeps the evidence attached.

CurrentTech Ops System Administrator at Avelo Airlines

FounderAngel Software Solutions LLC

FocusEmerging technology evaluation and agent systems

Selected systems

Different projects. Different engineering problems.

AgentReef coordinates work between agents. Kynetik turns visual ideas into working sites and readable code. Arena tests bots in motion. Aquarium tests what a browser can carry. Each one asks a different engineering question.

AgentReef homepage showing the agent federation platform Active development
01 / Agent infrastructure

AgentReef

AgentReef is the coordination layer I wanted when several agents were working on one codebase. Peers can discover each other, call skills, share signed memory, and leave a durable record of the work.

The server is written in Rust and uses signed requests, scoped grants, PostgreSQL row-level security, and Redis for presence. Most of the engineering work is in keeping ownership, context, and permissions clear after the work spreads out.

  • Rust API and WebSocket server
  • Ed25519-signed calls and results
  • Shared memory with tenant isolation
  • Operator dashboard and audit trail
Kynetik website builder showing its visual editing workspace Visual product platform
02 / Product engineering

Kynetik

Kynetik is the website builder I wanted for people who need real control without starting from a blank code editor. It combines visual editing with responsive previews, reusable content, collaboration roles, export, and feature entitlements.

Trigger Studio adds visual behavior tools, a constrained interpreter, linting, preview, and readable HTML, CSS, JavaScript, and PHP output. A Fastify and PostgreSQL backend handles authentication, access control, rate limits, and API keys.

  • React and TypeScript builder
  • Visual behavior and code tools
  • Fastify and PostgreSQL backend
  • 23 integration tests
AI Battle Arena interface showing its bot combat platform Open source / AGPL-3.0
03 / Real-time systems

AI Battle Arena

Arena is a live combat sandbox for AI bots. Authors deploy a bot, watch a match, and iterate against a scoreboard that keeps moving.

A Go server owns the match state. WebSockets stream combat to the browser spectator UI, while Python and Node SDKs give bot authors a practical way in. The public repository includes the protocol, examples, and contribution guide.

  • Go server and WebSocket protocol
  • Browser spectator interface
  • Python and Node SDKs
  • Public AGPL-licensed repository
Aquarium 3D underwater research station running in a browser 3D browser experience
04 / Visual computing

Aquarium

Aquarium is a cinematic underwater world that runs in the browser. It has a research station, schools of fish, ambient sound, changing light, and enough movement to make performance work matter.

The current experience runs on Babylon.js and WebGL. I also wrote the clean-room rebuild plan for a WebGPU-first version with deterministic simulation, adaptive quality, safe teardown, and accessibility built into the architecture.

  • Procedural 3D world and creature simulation
  • Babylon.js and WebGL
  • Adaptive rendering and reduced-motion planning
  • WebGPU-first rebuild specification

Operating background

The AI work grew out of systems people already rely on.

Airline records and software releases leave little room for vague ownership. That work taught me to care about the source, the handoff, and what happens when the normal path breaks.

Open the full resume
  1. Aug 2025 - present

    Tech Ops System Administrator

    Avelo Airlines

    Support AMOS and workflows used by Maintenance, Aircraft Records, Planning, and Operations. Trace access, data, and process issues through the source, the handoff, and the owner responsible for the next action.

  2. Jan 2026 - present

    Founder and Operator

    Angel Software Solutions LLC

    Set product direction, technical priorities, release scope, and deployment standards across AgentReef, Kynetik, AI Battle Arena, Aquarium, and Angel-Serv.

  3. Jun 2024 - Aug 2025

    Aircraft Records / Records Technician

    CommuteAir

    Reviewed aircraft maintenance records under FAA requirements, entered maintenance actions in AMOS, and checked records for completeness, legibility, and traceability.

  4. Jan 2023 - Jun 2023

    Release Validation, First Customer Ship

    Apple

    Validated more than 10 OTA software updates with manual checks, automated scripts, defect reports, retesting, and release review.

  5. Apr 2022 - Jun 2024

    Technical Expert

    Apple

    Diagnosed device and software issues, repaired more than 200 devices in a typical month, maintained a 99% repair resolution rate, and trained teammates.

  6. 2010 - 2018

    Founder and operator

    Angel-Gaming

    Ran eight global game servers and led more than 20 staff across seven time zones. The community grew to more than 20,000 members.

Open source and community

I like the part where people have to understand each other.

I joined the OpenClaw community in January 2026 and became a Member of the Community Staff in March. I answer questions, keep discussions useful, surface issues, and help people find enough context to move forward.

That work has made me better at the less visible side of engineering: explaining a decision, noticing where documentation failed, and keeping a technical conversation from losing the people it is meant to help.

Visit OpenClaw
A useful thread usually needs
01

The real questionWhat is the person actually trying to make work?

02

Enough contextWhat changed, and what has already been tried?

03

An ownerWho can answer, test, or make the next decision?

04

A follow-upDid the answer solve the problem, or only sound right?

Working set

Tools I use in the work

AI and agents

Claude and Claude Code, OpenAI ChatGPT, GPT and Codex, GitHub Copilot, Gemini, OpenClaw

Software

Rust, Go, Python, JavaScript, TypeScript, SQL, React, Fastify, REST APIs, WebSockets

Infrastructure

PostgreSQL, Redis, Linux, Docker, Cloudflare, GitHub Actions, authentication and access control

Operations and evaluation

AMOS, Jira, Bugzilla, Splunk, Tableau, requirements, pilots, test cases, release validation

What I am looking for

Work where AI has to fit inside a real operation.

I am looking for emerging technology, agentic AI, technology operations, or technical product work. Aviation is a natural fit, but I am interested anywhere software needs clear ownership, useful evaluation, and people who can explain the next decision.