AI/ML Engineer
Building the infrastructure that runs AI in production.
AI/ML Engineer Co-founder LAB32.AI Building AI infrastructure 32× RTX 3090 GPU cluster
LAB32 compute
32 × NVIDIA RTX 3090
- GPUs
- 32
- Total VRAM
- 768 GB
- CUDA cores
- 335,872
- Workloads
- Training, tuning, serving
The first 32× RTX 3090 cluster of its kind — InfiniBand interconnect between every node, and the whole rack runs on solar.
- InfiniBand fabric
- 100% solar powered
- Built in-house
About
From backend systems to GPU racks.
I'm an AI/ML engineer and co-founder of LAB32.AI, where we take AI past the demo stage — fine-tuning and serving models, building RAG and agent systems, and running the infrastructure underneath them so they hold up under real traffic.
A decade of Python, Go, microservices and cloud infrastructure sits underneath all of it. Today that experience goes into private deployments companies can actually own, on 32 RTX 3090s we built and run ourselves.
Most of what I learn ends up as open-source tooling or notes on the blog. Models and datasets land on Hugging Face.
- Fine-tuning
- Inference serving
- RAG & agents
- GPU cluster ops
- Python
- Go
Open source
Tools built for the cluster.
nvidia-fan-control
Keeps fan curves under control on dense multi-GPU Linux boxes, so cards stay in their thermal envelope.
View on GitHub →gpu-stats
A small Go service that exposes GPU utilization, memory and temperature as JSON for dashboards and alerts.
View on GitHub →sshmenu
A terminal picker for jumping between cluster nodes without memorizing hostnames.
View on GitHub →Tools
Interactive references I built for my own work: a fine-tuning method explorer and an inference engine explorer, both checked against primary sources.
Open the tools →Hugging Face
Models and datasets I publish from experiments run on the LAB32 cluster.
View profile →Get in touch
Taking AI into production?
Private model deployments, fine-tuning on your own data, RAG and agent systems, or infrastructure that has outgrown a single machine — happy to talk.