Columbia University

Technology Ventures

ZSS Microsystems

ZSS Microsystems: Our technology uses mixed signal circuitry to perform the main computations of deep learning within the SRAM block. This technology will serve to accelerate machine learning at the edge for the ever-growing internet-of-things market such as UAV and wearables. Machine learning’s main computation primitive is multiply and accumulate (MAC). Algorithmic research has been able to create methodology which can reduce MAC computation to binary while maintaining accuracy for many real world metrics. Our technology utilizes capacitive coupling to perform binary MAC in the analog domain which substantially increases the throughput and lowers the energy cost. The technology also uses capacitor based unit of charge which is superior to current based computing method for having lower variability. The in-memory-computing module is integrated with an easy-to-integrate microcontroller such that it can be effortless be utilized in consumer IoT products.

To learn more about this opportunity, contact:

Greg Maskel
Technology Licensing Officer
gregory.maskel@columbia.edu