Biography
Pedro Cuenca is a machine learning engineer at Hugging Face, where he works on open-source software, developer advocacy, and on-device AI. He was previously a co-founder of LateNiteSoft and worked on Camera+, a top photography app for iOS and one of the first to introduce custom local models for photography enhancement, spawning his strong interest in local AI. He was also involved in the development of and operations behind DALL·E mini. Pedro is based in Spain, collects big GPUS (less than he'd like) and small single-board computers (more than he needs). He contributes to multiple open-source repos and likes to share insights on open models, especially if they can be optimized to run locally.
Session
- AI For TechSpeaker
Workshop 15 - Not your weights, not your brain!
Wednesday 23 September, 13:30 – 14:50Workshop Room ARoom full? This workshop is also streamed live in Workshop Room B at the same time. Can open models drive your agentic needs? This session explores the current state of open AI, including local AI. It then goes on to examine how to incorporate open models to your workflow. Open models come in many flavors and sizes: from large near-frontier behemoth models, to small models that can run on your laptop. We'll show when best to use each, and how to integrate them with your favorite tools for conversational and agentic use cases. We'll further explore how quantization libraries such as llama.cpp and MLX allow large models to run locally, achieving more control, total privacy, low latency, no cloud costs.