Specializing in agent memory systems, orchestration tooling, and production-ready AI deployment. Everything I ship runs locally, scales efficiently, and respects your data.
Selected Work
Open tools for AI developers — focused on memory, deployment, and agent orchestration.
Agent memory system that reduces context token usage by 70–98% without embeddings or GPU.
Uses FTS5 indexing + LYT knowledge mapping
View projectAutomated memory management for AI agents. Tier-based retention, FTS5 search, and Obsidian sync — no GPU required.
SQLite + Python • Runs on $5/mo VPS
Explore CORTEXOne-command VPS installer for OpenClaw. Provisions and secures a production AI gateway in under 5 minutes.
Supports SumoPod, Alibaba Cloud, and generic VPS
ESP32-S3 4WD companion robot with BLE control, computer vision, and direct LLM brain. Embodied AI at the edge.
ESP32-S3 • BLE • Computer Vision
Meet KodaPersonal knowledge management with ACE framework and AI-generated Maps of Content. Automated nightly synthesis.
Obsidian-based • AI-powered MOCs
Explore systemPractitioner's guide to running a self-hosted AI gateway reliably — deployment, recovery, and hardening.
For OpenClaw operators
Read guideEarly experiment in GPU-based codebase memory via embeddings. Superseded by graymatter's lighter, CPU-only approach.
Archived — see graymatter for current implementation
View archiveGenerative Media
Experiments in AI video generation using Veo 2, custom TTS pipelines, and automated post-production.