Participated · Series B · Dec 2020

AIStorm is a pioneer of AI-in-Sensor processing that accepts charge directly from sensors (pixels or MEMS microphones) and performs charge-domain analog computation to run convolutional neural networks at the sensor. Its charge-switched analog approach multiplies charge directly, enabling MAC efficiencies up to several thousand TOPS per watt and minimizing the need for external components. The technology targets ultra-low-power, ultra-low-latency always-on edge use cases across imaging and audio, including facial recognition, access control, behavioral tracking, people counting, object tracking, and segmentation. AIStorm positions its solution for battery-operated devices such as cell phones, TV remotes, laptops, home security, and authentication systems. To accelerate commercialization and market access, AIStorm has strategic partnerships with AsusTek, Egis Technology, Knowles Corporation, and Meyer Corporation. The company is headquartered in Houston, Texas, with offices in Graz, Budapest and Hsinchu. AIStorm is a pioneer in high-performance AI-in-sensor processing that operates on sensor data in its native analog form to avoid digitization. Its approach enables real-time edge processing, reducing latency and power consumption while claiming five- to ten-fold cost savings versus GPUs. The technology targets applications where fast reaction and low power are critical, with explicit relevance to advanced driver-assistance systems and mobile devices. Company executives say the architecture allows intelligent pruning of sensor data streams in real time, keeping up with massive sensor inputs. AIStorm is headquartered in Silicon Valley with offices in Phoenix and Graz, Austria, and its team includes industry veterans from leading semiconductor firms. The company recently closed a $13.2 million Series A to accelerate engineering and go-to-market efforts.

Led · Equity · Jun 2020

Deep01 builds AI software intended for emergency departments to help physicians interpret CT brain scans faster. Its software can detect acute intracerebral hemorrhage with a reported accuracy of 93% to 95% and returns results within about 30 seconds. The product has obtained clearance from both Taiwan and the U.S. Food and Drug Administrations. Deep01's software is deployed in two medical centers and four hospitals in Taiwan and has helped clinicians review more than 2,000 brain scans. The company received its first purchase order, worth about $700,000, in February. Deep01 was launched in 2016 by founder and CEO David Chou, and a partnership with ASUSTek is described as promising for smart medical applications.