Doing big things with "small" hardware. 🚀
I’m an AI developer focused on the 1.5B – 4B parameter sweet spot. While others throw clusters at a problem, I fine-tune on an iGPU 780M using ROCm 7.2.1, Unsloth, and QLoRA to prove that efficiency is the ultimate scale.
Capabilities:
Data: Custom curation & synthetic dataset generation.
Training: Highly optimized QLoRA adapters.
Operations: End-to-end pipelines from raw text to monitored endpoints.
Check out my models below to see what localized compute can really do.