TCH AI MDA SA Samuel Adeyeye Scaling Qwen2.5‑VL Fine‑Tuning with DeepSpeed ZeRO‑3 Across Two DGX Sparks (Docker, Detached) Fine‑tuning a 32B‑parameter vision‑language model is where “it works on my laptop” stops being a joke and starts being an outage risk. In…
ART HUM AI SA Samuel Adeyeye Teaching a Vision Model to Read Retail Leaflets: How We Built Production VLM Inference A practical architecture for document-style vision extraction — without treating the model as your business logic engine.
