Randy Johnson
I make AI and Python do things so you don't have to
New to Contra
Randy is ready for their next project!
Split-Compute Stack
I set up a home AI lab with my Mac M4 handling LLM inference via LM Studio and my Windows desktop with an RTX 3060 handling image generation. They're on a dedicated subnet behind an OpenWrt router, meshed together with Tailscale so I can access everything remotely from my phone too.
1
9
ComfyUI Workflow Automation
I use ComfyUI a lot for AI image and video generation and ended up building a Python layer to automate and control it programmatically. This particular workflow is a stop-motion video pipeline using Wan2 with 31 nodes. I've also built LoRA training pipelines and batch processing tools on top of it.
1
17
Raphael
I built Raphael as a personal experiment in what it would look like if an AI system could manage itself like an operating system. It coordinates 24 specialized agents, handles its own memory via Neo4j, and has a live telemetry dashboard. It's probably the most ambitious thing I've built and it's still evolving.
1
27
Memory Graph Library
I got frustrated with how most RAG setups lose context over time, so I built my own memory layer on top of Neo4j that combines graph relationships, vector search, and fulltext retrieval. It's designed to give AI agents actual persistent memory that scales across multiple tenants.
1
19