Evidence over promises.
This page distinguishes clearly between client work, research, open-source engineering, and experiments. We publish client results only when they are real and permitted — never invented logos, quotes or statistics.
Strix Halo Local AI Guide
Open research on running large language models on your own hardware.
An independent, practical, evidence-led guide to deploying and benchmarking large language models locally on AMD Ryzen AI MAX+ 395 / Strix Halo systems. The project covers local model serving, Ubuntu configuration, Ollama, llama.cpp, Vulkan/RADV, model selection, benchmarks, raw evidence, and reproducibility.
- documentation
- benchmarks
- reproducibility
- community-findings
- local-inference
What it proves about us: hands-on local AI research, hardware-aware model deployment, reproducible benchmarking, open technical documentation, and a practical understanding of private inference.
Not a customer case study, not an official AMD project or partnership, and not a promise that every model or workload achieves the same performance. Results are specific to the tested hardware, models and configurations.
- README.md — deployment guide
- benchmarks/ — raw evidence
- ubuntu-setup.md — configuration
- model-selection.md — Ollama · llama.cpp
- vulkan-radv.md — GPU acceleration
Case studies appear here
when they are real.
Our policy is strict: real company, real process, real measured outcome, published with written permission. Until a project clears that bar, this section stays honest — and empty.
- Technical articles
- Architecture explorations
- Open-source engineering
- Experiments (clearly labeled)
Want to be the first case study we publish?
Early clients get senior attention and honest engineering. Bring a process worth automating.