Taming the AI Hardware Cambrian Explosion — Abdul Dakkak, Modular

Why watch

Abdul Dakkak explains how the proliferation of AI hardware created a fragmented software stack and how Modular's Mojo language and MAX inference framework address this. Covers benchmarks against vLLM, torch.compile, and FlashAttention 4, with sub-second image generation demos. For AI engineers, ML infrastructure developers, and systems programmers building high-performance inference pipelines.