Founder · Systems Builder · Linux Operator · AI Infrastructure Engineer
I build the whole machine—from bare metal and private cloud to local models, agent orchestration, production software, and the business systems around it.

Most developer profiles are inventories of tools. Mine is a map of systems I have built, integrated, deployed, broken, repaired, secured, monitored, and learned well enough to own.

I am the founder and CEO of MADPANDA3D, a veteran-owned technology company. My work crosses hardware, Linux, private cloud, local AI, agentic automation, full-stack product engineering, observability, cybersecurity, and business operations. I am comfortable moving from a rack and an HBA to a kernel log, from a model router to a React interface, and from a customer problem to the production system that solves it.

My lane is the space between disciplines—the part where somebody has to understand how all of the pieces become one dependable operating system.

leo@madpanda:~$ whoami
Leo Lara — founder, builder, operator, and relentless systems student.

leo@madpanda:~$ mission
Own the stack. Reduce hidden dependencies. Automate with accountability.
Build useful things. Leave proof.

My career has repeatedly placed me where failure has a real cost.

• United States Marine Corps: four years as an Aircraft Rescue Firefighter with MWSS-172 in Okinawa—training, readiness, teamwork, and calm execution when the situation is not theoretical.
• Telecommunications: Tiger Team field work supporting T-Mobile across Missouri, Illinois, and St. Louis—network troubleshooting, emergency response, break/fix repair, and restoring service under pressure.
• Cloud engineering: nearly two years in an AWS engineering environment in Columbus, learning how infrastructure, scale, process, and operational discipline meet.
• MADPANDA3D: founded in January 2024 around 3D printing and design, then expanded into software, AI infrastructure, Linux, automation, websites, private cloud, and business systems.
• Cybersecurity: currently deepening the security side of the stack through formal Cybersecurity studies at National University and continuous hands-on lab work.

That history is why I do not separate software from operations, security from usability, or automation from responsibility.

flowchart TB
    Leo["Leo / Human authority"] --> O["Odysseus / Authenticated orchestration plane"]

    subgraph AI["Self-hosted AI and agent systems"]
        O --> J["JARVIS / Warm local Qwen brain on Ollama"]
        O --> W["Workstation / Whisper + Chatterbox + Codex"]
        O --> H["Gordon / Governed business operator"]
        O --> K["Knowledge OS / Chroma + FastEmbed + Obsidian"]
    end

    subgraph CLOUD["Private cloud"]
        N["Nimbus / Proxmox"] --> T["TrueNAS / HBA-backed storage"]
        T --> I["Immich"]
        T --> C["Nextcloud"]
    end

    subgraph PRODUCTS["Products and control surfaces"]
        P1["Charter"]
        P2["MAD Card Exchange"]
        P3["MAD MCP Portal"]
        P4["SOC / NOC Wallboard"]
        P5["MADPANDA3D.com"]
    end

    OBS["Grafana · Prometheus · Loki · Alertmanager · Alloy"] -. observability .-> AI
    OBS -. observability .-> CLOUD
    O --> PRODUCTS

This is not a diagram I made for a portfolio. It is an active, evolving home-lab environment with dedicated compute roles, storage, voice, retrieval, orchestration, security boundaries, and observability.

Nimbus is a working local-first cloud baseline under active reliability hardening. I document failure modes and revision paths instead of overstating readiness.

voice or chat
    ↓
Whisper speech recognition on the workstation
    ↓
Odysseus authentication, session state, policy, and routing
    ↓
Qwen 3.5 on the dedicated JARVIS/Ollama compute node
    ↓
governed local tools, Codex work, or approval-aware agent delegation
    ↓
Chatterbox Turbo speech synthesis with progressive PCM playback

The knowledge layer is filesystem-first: curated Obsidian/Markdown remains canonical, while Chroma and CPU FastEmbed provide derived retrieval. A July 2026 governed snapshot contained 2,781 sources and 10,849 retrieval chunks. The point is not the count; the point is provenance, review gates, reproducibility, and keeping derived memory subordinate to source truth.

I work across the full AI application path: model serving, speech, context, retrieval, tool protocols, agents, product interfaces, security boundaries, evaluation, and operations.

My current daily driver is EndeavourOS, built on the Arch ecosystem, running Hyprland on Wayland through a custom HyDE environment called MADPANDA Dark Zen.

I do not “rice” Linux and stop at screenshots. I debug compositor state, repair plugin ABI drift, write recovery services, build installers, test restore paths, and keep the environment reproducible.

• Custom EWW bars, widgets, menus, lock surfaces, and monitor-aware layouts
• Hyprland keybindings, workspace routing, hotplug/sleep recovery, and HyprPM plugin repair
• A materially patched Rust/GTK4 layer-shell dock with per-monitor workspace state
• AWWW wallpaper orchestration across mixed landscape and portrait displays
• Kitty, Zsh, Starship, Fastfetch, Dolphin, Rofi/Hyprlauncher, PipeWire, and Wayland-native utilities
• systemd services and timers for automation, backup, RGB, runtime repair, and health checks
• pacman/yay workflows, AUR inspection, ArchISO builds, reproducible installers, and rollback-aware package changes
• LUKS/LVM and LUKS/Btrfs systems, GRUB, dracut, Timeshift restore tooling, firewalld, and NVIDIA/Wayland diagnostics
• Docker GPU services, Proxmox VMs/LXCs, TrueNAS Apps, and hardened Ubuntu workers

A real shout-out to the EndeavourOS, Arch, and Hyprland communities. My workstation is proof that Linux can be a serious engineering platform, an operations console, and a deeply personal interface at the same time.

Public Linux work: ARCH-KITTY-CONFIG · HYPRLAND-THEMES

These repositories show different parts of the operating range—from AI memory and CRM automation to media pipelines and Linux desktop engineering.

MADPANDA3D is where the engineering meets real business operations. I work across discovery, architecture, implementation, deployment, client communication, automation, and ongoing ownership.

• Full-stack websites, SaaS products, portals, dashboards, and internal tools
• GoHighLevel CRM architecture, lead pipelines, contacts, opportunities, messaging, calendars, invoicing, and reporting
• n8n and API-driven business automation
• SEO, AEO/GEO, content systems, funnels, analytics, and conversion infrastructure
• Client onboarding, operating ledgers, approval workflows, recurring reporting, and evidence-backed delivery
• AI-assisted content and operations with human review gates
• VPS, Docker, Nginx, Cloudflare, DNS, certificates, private networking, and deployment runbooks
• Brand systems, 3D design, media pipelines, and marketing assets when the product needs more than code

The business lesson is the same as the engineering lesson: a tool is not a system until ownership, state, failure modes, and the next responsible action are clear.

1. Start with source truth. A dashboard is a view; the ledger, runtime, filesystem, or authoritative API owns the state.
2. Make dangerous actions deliberate. Read-only by default, dry-run when possible, confirmation and scope when mutation is necessary.
3. Build rollback before bravado. Backups, manifests, bounded changes, validation, and a known way home.
4. Treat observability as part of the product. Health, events, logs, evidence, and failure states should be visible.
5. Use AI as force multiplication, not borrowed credibility. Agents can research, implement, test, and review; I remain responsible for architecture, permissions, truth, and the result.
6. Own the handoff. The work is not finished until another human—or tomorrow's version of me—can understand and operate it.

• Bringing Charter through launch-stage acceptance and operational hardening
• Advancing the JARVIS + Odysseus voice-first, concurrent-agent operating model
• Expanding governed MCP infrastructure and the MAD MCP Portal
• Hardening Nimbus private-cloud reliability, backup, and observability
• Building Gordon into a safe, accountable business operator
• Continuing formal cybersecurity education and applying it to real systems
• Helping veteran-owned and small businesses replace scattered tools with one practical operating stack

I am not claiming to know every language, every model, or every layer better than a lifelong specialist in that one layer.

I am claiming something different—and I have the systems to prove it:

I can take an idea from the physical machine to the operating system, from the model and data layer to the interface, from deployment to observability, and from technical capability to business value—without losing responsibility between the layers.

If your organization needs somebody who can speak with leadership, inspect a kernel log, architect an agent system, ship the product, deploy it, instrument it, and explain the risks in plain language, we should talk.

LeoLara@madpanda3d.com  ·  madpanda3d.com  ·  Discord  ·  LinkedIn

     .+######+            .++#####+.
   +###+....+##+++++.+++++##+....+###.
  +##+     .+##############+.      +##.
  ##+    .+###################+     ##+
  ###.  +#######################.  .##.      ##          ##  ##########  ##########  ########
  .+##.+#########################+.##+       ####      ####  ##      ##          ##  ##     ##
    ..+###+.++#############+.++###+.         ## ##    ## ##  ##      ##         ##   ##      ##
     +##+      +########+.     .+##+         ##  ##  ##  ##  ##########      #####   ##      ##
    ###+    @@@  .####+.  @@@    +##+        ##   ####   ##  ##                 ##   ##      ##
   .###.     @@. .####+. @@@     +###        ##    ##    ##  ##                  ##  ##     ##
   .###+       .+#######+       .###+        ##          ##  ##          ##########  ########
    .####+    .##++##++###.    +###+.
     .+####+. .##+.   .+##  .+###+.
        .++##+.+###++####..+#++.
             .. .++###++  ..

SYSTEM STATUS: BUILDING
CONTROL MODE: HUMAN-IN-THE-LOOP
DEFAULT POSTURE: PRIVATE · OBSERVABLE · REVERSIBLE

_{Built by Leo Lara in South Florida. Marine veteran. Founder of MADPANDA3D. Still learning. Still shipping.}