Ludwig Computing creates next-gen computing platforms with a revolutionary approach. Avoiding traditional clocked digital platforms, Ludwig’s paradigm enables a natural computing flow via algorithm-hardware codesign, harnessing the natural physics of devices for Generative AI and probabilistic computations, significantly reducing computation time and cost for these problem domains.
FELLOWS
Behtash Behin-Aein
Behtash Behin-Aein received his Ph.D. in electrical and computer engineering from Purdue University in 2010. His spintronics research was one of the rare theoretical publications in Nature Nanotechnology. Later, at GLOBALFOUNDRIES, Behin-Aein led the development of device and compact models for STT-MRAM chips from concept to manufacturing 40Mb arrays. The decade-long pursuit of his passion, probabilistic computing, has led him to co-found a startup to actualize a new computing platform.
Jan Kaiser
Jan Kaiser is a co-founder of Ludwig Computing. He earned his Ph.D. in electrical and computer engineering as a Ross fellow in Supriyo Datta’s lab at Purdue University. His research focused on spintronic devices and probabilistic computing systems. He is passionate about novel physics-inspired computing approaches. Kaiser earned bachelor's and master’s degrees in electrical engineering from Ruhr-University Bochum in Germany.
TECHNOLOGY
Critical Need
Data centers and information and communication technologies, including personal devices such as computers and mobile phones, are on track to consume more than 20 percent of global electricity by 2030. Plus, scaling today’s computers based on Moore’s law presents daunting challenges, limiting performance improvements with reasonable costs.
At the same time, many industries are looking to probabilistic computing, which uses statistical methods as a means of solving problems, to improve their businesses. Running probabilistic computing on general-purpose microprocessors, however, means that the benefits it offers to industry come at the expense of increased silicon footprint and energy consumption. Efficient domain-specific accelerators can rectify this problem.
Technology Vision
Probabilistic computing makes it possible to process uncertainties inherent in data or leverage randomness to interpret, infer, and make better decisions faster and more efficiently than conventional computing.
Probabilistic co-processors are domain-specific accelerators designed to execute probabilistic computing algorithms faster while consuming less space and energy than conventional general-purpose chips.
Rather than shrinking devices to pack more into a given chip area, Ludwig Computing is reducing the number of devices and simplifying the circuits by changing the information processing paradigm to increase computational power.
Ludwig Computing is developing its probabilistic co-processors for applications in Generative AI and complex optimization problems.
Potential for Impact
This technology could create a new paradigm for faster and more efficient implementations of computationally intensive probabilistic algorithms. It will enable the computations on wearable devices and at the edge instead of in the cloud. It will reduce latency and allow real-time applications either not possible or too cost/power prohibitive today.
Probabilistic computing can improve compute performance without increasing transistor count—an increasingly complex task—and can usher in a new era of progress.