Vierla

Independent AI research & development, from training efficiency to running large models on the hardware people actually own.

Research areas

• Efficient inference: GPU memory tiering (VRAM → system RAM → disk), CUDA VMM, 70B+ models on consumer GPUs
• Training efficiency: optimizers and speedruns (Muon, modded-nanogpt)
• 3D & vision: Gaussian splatting, optimal-transport morphs, vision foundation models
• Agent infrastructure: MCP tooling, local control planes for coding agents

Models, datasets, and demos will be published here as they're ready.

🌐 vierla.com · 🧑‍💻 AmeerJ97 on GitHub