First Local Model
Establish a baseline by running a local language model on the Tesla P100 and documenting setup, compatibility, and performance.
EXPERIMENTS
Testing what the hardware can actually do.
This is where theory meets power consumption, driver issues, memory limits, model compatibility, and real-world performance.
GPU Validation
Verify GPU detection, driver installation, CUDA availability, memory reporting, and sustained operation under load.
Performance Testing
Compare model sizes, quantization levels, response speed, memory usage, and power behavior.
Building the Platform
Document the infrastructure work required to make private AI practical, maintainable, and accessible from the network.