TAKTICAL TAKO / TECHNICAL PROFILE

Bryan

Linux · Networking · Local AI · Self-hosted systems

takticaltako.com · Project portfolio · October 2026

Hands-on homelab work spanning Ubuntu server administration, domain routing, local inference, and event-driven software. Interested in infrastructure and the practical integration of AI into working systems.

Technical experience through projects

Systems: Ubuntu, SSH, systemd service management, Docker, Nginx, and Samba.
Networking: Cloudflare DNS, locally managed tunnels, hostname-based routing, and loopback service binding.
Local AI: Ollama, quantized GGUF models, a resident Python/PyTorch inference worker, deadlines, and fallback behavior.
Software: Node.js, Java/Paper plugin projects, Python, SQLite, REST APIs, and Discord integration.

Selected projects

Self-hosted portfolio infrastructure

Hosts this domain through Cloudflare Tunnel and Nginx on an Ubuntu homelab server. Dedicated origin routing preserves other hosted applications. Service boot configuration, configuration validation, backups, and installation rollback support repeatable deployment.

Pantheon — local AI decision director

Experimental Minecraft integration pairing a Paper plugin with a Node.js director and resident Laya inference worker. Uses prepared actions, response validation, bounded concurrency, and cosmetic fallback. Smoke evaluations surfaced instruction-injection and option-order failures, motivating a cosmetic-only model rollout. Local Laya service verified running; broader gameplay validation remains separate.

AMP playtime dashboard

Project implementation converts game-server logs into SQLite-backed player sessions and exposes them through a dashboard, REST API, and Discord commands. Handles persisted file offsets, ordered backfill, failed handshakes, stable player IDs, and shutdown events. Source inspected; current deployment not verified.

Local model hosting

Runs a loopback-bound Ollama service with a quantized MiniCPM model installed. Explores inference on shared homelab hardware, separately from the Pantheon Laya runtime. No inference performance or model quality claim is made.

Lab environment

AMD Ryzen 7 5800X · 8 cores / 16 threads · 32 GB RAM · Ubuntu. Runs web infrastructure, local AI services, and hosted game workloads.

This profile summarizes personal projects, rather than employment history or certifications. Contact details are available in the application accompanying this portfolio.