Architectural Portfolio
NemoClaw Agent Watchdog & Security Sandbox
A zero-dependency health watchdog, incident logging daemon, and Landlock/Capsh container security sandbox for multi-agent LLM runtimes (OpenClaw, Hermes). Monitors container process health and dispatches instant telemetry to Discord and ntfy.sh.
OnboardBob
An autonomous codebase topology cartographer and onboarding copilot built with 3 parallel subagents (Env Auditor, Repo Mapper, Runtime Checker). Features a 3-layer prompt injection defense against poisoned repositories.
Research Navigator
High-level system design case study of an automated validation platform. Orchestrates a high-performance Go WebSocket proxy runtime, Python LLM-as-a-Judge audit engine, and dynamic user dashboard.
MyBriefcase Legal Suite
A court-compliant document bundling tool. Aggregates multiple legal PDFs, generates automated cover pages and indices, and uploads compiled briefs to AWS S3.
Newsletter Herald
An automated document processing pipeline. Parses incoming PDF and Word newsletters, standardizes metadata, generates multi-stage AI summaries, and dispatches weekly newsletters to a parish subscriber directory.
PulseApp (Yello Promises)
A peer-to-peer appreciation feed. Implements points allocation rules and real-time activity updates with Azure Blob file attachments.
Legal Document Sorter
A local background daemon that monitors download directories, uses regular expressions to classify legal documents by case caption, and automatically synchronizes them to Google Drive and OneDrive folders.
Technical Architecture Notes
Hardening the Fleet: The Battle-Tested Pitfalls & Gotchas of Multi-Agent Linux Sandboxes
Taking The Claw Deck from a single test container to a full 3-agent autonomous fleet (Hortense, Gwen, and Makeda) with Discord, Tailscale MagicDNS HTTPS, and live read-write Obsidian/Google Drive sync. Here are the 8 real-world pitfalls, proxy drops, POSIX ACL collisions, and Docker gotchas you only discover in the trenches.
2026-08-25Spawning The Claw Deck: Running Autonomous AI Agents on Fedora Linux
Building a multi-agent headquarters on Linux. Setting up our first sandbox agent Hortense with Hermes and MiniMax M3 via NVIDIA NIM, debugging Docker permissions and OpenShell gateway version drift, and creating The Claw Deck command center in Discord.
2026-08-24Tamriel on Fedora: Conquering the 100GB Linux Gaming Hurdle
The ultimate test of wiping Windows was whether my favorite MMO—The Elder Scrolls Online—would actually run. From 2023 installer timeouts and host_ieversionfail registry traps to unlocking a 240Hz screen: here is how we got ESO running at full speed on Fedora 44 with an RTX 4070.
2026-08-23Going Cold Turkey: Why I Finally Wiped Windows on My Main Machine
After years of bouncing off Ubuntu, Mint, and Debian at the first sign of trouble, I changed my strategy. From a headless HP homeserver with CasaOS to wiping Windows 11 on my Razer Blade 14 (RTX 4070)—here is my honest Day 1 reality check.
2026-07-15Architecting Multi-Agent AI Pipelines with Go, Vue 3, and Python
How we built Research Navigator using a high-performance Go proxy runtime, a Vue 3 Traffic Light UI, and an asynchronous Python LLM-as-a-Judge audit engine.
2026-07-02Resolving Channel Failures in Containerized AI Agent Deployments
A deep dive into isolating agent runtimes using Docker, mapping internal gateway ports, and ensuring robust failover over private Tailscale networks.