Max Zanoga is an AI/ML engineer and co-founder of LAB32.AI. He builds private model deployments and operates a self-built cluster of 32 RTX 3090 GPUs.[1]
Start with AI infrastructure tools →
Links
Max Zanoga: personal site Personal site
GitHub 
X LinkedIn Hugging Face Co-founder and AI/ML engineer · LAB32.AI
Career
Zanoga describes a background in Python, Go, microservices and cloud infrastructure. At LAB32.AI he applies that experience to fine-tuning, model serving, retrieval systems and agents. The practical aim is to put models on infrastructure that a company can control, rather than leave the work at a prototype. His personal site says the cluster connects nodes with InfiniBand and runs on solar power. These are his descriptions of the installation, rather than independently measured efficiency claims.[1]
Contributions
His interactive references turn recurring infrastructure choices into tools. The fine-tuning explorer compares methods and narrows choices by hardware constraints. The inference explorer compares serving engines. A VRAM calculator separates model weights, the cache used during generation, and training optimizer state. That breakdown helps explain why a model that fits for inference may need much more memory to train, or why a longer context changes the deployment budget.[3]
Tools, writing & talks
Apps
Start here
AI infrastructure toolszanmax.github.ioInteractive guides for choosing fine-tuning methods, inference engines, GPU hardware and quantization formats, plus a model-memory calculator.
Start with the VRAM calculator in this collection if you are choosing hardware. Bring a model size, precision and expected context length; the component breakdown makes the tradeoffs visible.
Author[3]
Tools
- nvidia-fan-controlgithub.com
Linux tooling for controlling NVIDIA GPU fans, a practical concern when several cards share a dense machine.
Repository owner and developer[4]
Writing, research & talks
A first-person narrative article documenting the journey of scaling a personal GPU cluster from a single RTX 3090 to a 32-GPU setup over two years. It covers the practical decisions, challenges, and lessons learned in building serious personal AI/ML infrastructure. Useful for ML practitioners and hobbyists considering scaling their own compute.
Catalog workOriginal source