DD-AIM Information

The Company has invented and patented an innovative digital circuit design for a real-time AI/ML inference accelerator chip to be used in Predictive AI. What sets us apart: • Our scale allows running AI deep learning models incorporating new algorithms for increased forecast accuracy. • Our chip has far lower unit manufacturing costs, plus dramatically less power consumption and heat generation. • Even when implemented on an ASIC, model parameters and run-time configurations can be modified during inference. • The chip features fault detection and fault tolerant operations, plus self-learning, self-correcting, and self-optimizing routines via in situ retraining. We are building fully working production-ready example models in four targeted industry sectors to prove the practicality of our technology. • Defense: The combination of low power consumption/low heat generation with small size/low weight in a chip package offering extremely minimal unit costs along with in-the-field adaptability solves major problems for the DoD, with its emphasis on attritable hardware. Our hardware’s ability to run huge ensembles of neural networks in parallel allows remarkably effective implementations of cognitive sensor systems. • Finance: Quant-focused investment management and trading firms will be able to dramatically improve and scale their current AI/ML models to profit from opportunities missed now because of hardware and software constraints. • Healthcare: The ability to run a patient-specific auto-tuned AI/ML model for every patient being monitored is a game changer, raising the standard of care. • Retail product forecasting: A retail chain stocking system might involve 20,000 SKUs, 2,000 stores, and daily sales data. Current AI/ML systems require that inference at these volumes be done at aggregate levels only, but aggregation dilutes accuracy. An array of our chips, however, can indeed handle this volume cost-effectively, improving accuracy.