Silicon-native builders.
Two engineers — one from the AI software stack, one from the semiconductor floor — converging on the same thesis: the memory wall is the bottleneck, and the only way past it is into the die.
Mo
Co-founder · Software & AI
Software engineer with deep roots in AI and data systems, bringing early LLM experience from the ground floor.
Mo is a software engineer whose career has spanned the full AI stack — from data pipelines and feature engineering to large language model research and deployment.
At BlackRock, he worked at the intersection of data infrastructure and AI, building systems that process and reason over massive financial datasets.
He was among the earliest engineers working with LLMs before they became mainstream, experimenting with transformer architectures and training techniques in the pre-GPT-3 era.
That early exposure to the raw mechanics of model inference — and the hardware constraints that shaped it — planted the seed for ondie.
Max
Co-founder · Hardware & Silicon
Engineer bridging PCB manufacturing, AI hardware, and Taiwanese semiconductor supply chains.
Max is an engineer with a hands-on background in PCB manufacturing and AI hardware design. His work sits at the physical layer — the traces, vias, and die layouts that determine whether a chip design actually works.
His time at National Tsing Hua University in Taiwan put him at the heart of the global semiconductor ecosystem. NTHU is one of Taiwan's top engineering universities, and the proximity to TSMC and the broader Hsinchu Science Park cluster gave him direct exposure to advanced fabrication processes.
He built relationships with Taiwanese semiconductor manufacturers and developed a practical understanding of how designs move from schematic to silicon — the exact knowledge needed to take ondie from FPGA prototypes to ASIC tapeout.
Max's combination of PCB-level manufacturing expertise and academic semiconductor exposure is what makes ondie's hardware roadmap credible.
The intersection that matters.
Building on-die AI requires someone who understands model architecture at the software level AND someone who understands how designs become silicon. We cover both.
AI depth
Early LLM experience, production AI infrastructure at BlackRock, and a clear-eyed view of where GPU inference breaks down. Mo knows the model side — architecture, attention patterns, and the weight-loading bottleneck that makes GPU inference fundamentally limited.
Silicon access
PCB manufacturing, FPGA prototyping, and direct relationships with Taiwanese semiconductor fabs from NTHU. Max knows how to get from schematic to silicon — the fabrication realities that determine whether a chip design lives or dies.