Democratizing Frontier LLMs - Cloud-cluster scale intelligence running on any Desktop PC
Explore the engineering breakthroughs behind running state-of-the-art, ultra-large Mixture-of-Experts models (100B-600B+ parameters) entirely offline on consumer-grade desktop hardware. This session dives deep into our novel storage-centric inference engine, developed by Monadd-AI and a research partner, that leverages dynamic expert offloading, hardware-topology-aware scheduling and prefetching algorithms to run Frontier LLMs directly on PCs. Learn how we run cloud-computing scale AI models without cloud-computing scale resources, just on a Desktop PC, with a live demonstration of a massive MoE model running locally.
Obadiah Pewee
Senior technology executive and aerospace engineer with 5+ years of leadership across AI/deep-tech startups, defence engineering, and venture-backed enterprises. Currently CEO of Monadd-AI, developing a world-first platform leveraging frontier LLMs for autonomous “virtual employee” capabilities. Accredited AICD director with a Graduate Certificate in Business (University of Melbourne) and Wharton certification in Strategic Management, bringing executive capabilities in fundraising, investor relations, strategic planning, and multi-disciplinary team leadership. Co-inventor on multiple US and Australian patents spanning hardware, software, and hybrid systems, with deep technical expertise in HPC, numerical simulation, and complex systems design—including 3+ years of advanced FEA at BAE Systems Australia and founding an aerospace startup pioneering hydrogen-powered rotorcraft.