We're building a RAG powered hardware advisor that helps people pick the right AMD GPU/accelerator for their ML workload. Answering questions like "does the MI300X fit my 70B model fine-tune?" with grounded answers cited from official AMD docs, instead of scattered PDFs and guesswork. We're also fine-tuning an open model to reason about tradeoffs, with the full pipeline (embeddings, fine-tuning, inference) running on ROCm.
.png&w=256&q=75)
.png&w=256&q=75)