Research

computing platforms

For the Student: For the first time since the inception of silicon revolution, modern computing is facing roadblocks traceable to fundamental constraints of nano and microelectronics. To support emerging technologies like artificial intelligence (AI), machine learning (ML) and internet-of-things (IoT), we are at a dire need to reinvent computing hardware. The technologies we have today are designed for precision but those do not capture ‘perception’ which is the goal of our future computers. As a result, to deal with the current data deluge, computers are becoming extremely energy inefficient. The power consumption at data centres is now a growing threat that can potentially lead to catastrophes like nationwide power cut.

At CeNSE we are building new generation of computers that will operate like a brain which is the intelligent most computing entity known to us capturing abilities like cognition and decision making. We are designing a brain like computing machine on a chip (for artificial intelligence) with an aim to offer at least a million-time gain in the energy-delay-product, compared to the modern computers (a measure of computing efficiency). This requires innovation at an interdisciplinary level involving material scientists, physicists, chemists, electrical engineers and computer scientists. Students from any discipline with passion to make disruptive innovations are welcome to join us.

Prof. Sreetosh Goswami is building a neuromorphic electronic chip based on molecular platforms. Prof. Shankar Kumar Selvaraja is making photonic neuromorphic chips and Prof. Pavan Nukala is designing inorganic materials for brain inspired computing.

For Industry: CeNSE has both infrastructure and the expertise to build nano devices for emerging computing platforms. Based on our expertise in materials fabrication, nano patterning, electrical, optical characterization, system on chip design and packaging we are able to develop indigenous computing platform in house that can be used for applications ranging from image processing, cybersecurity to bioinformatics. Few of the most promising areas we are working on are:

Analog computing, DNN: We are developing analog computing platforms using emerging organic and inorganic materials for implementing artificial neural networks and deep neural networks. Going beyond the existing state-of-the-art of 8-bit analog platforms, we are making >12-bit analog chips that could offer substantial efficiency in applications like image processing, recognition of speed of motion etc. We are employing combination of tabletop characterization tools as well as on-chip platforms to deliver these goals.

Spiking Neural Networks: Spiking neural network is the one that can be pushed on the verge of chaos. We are developing such platforms based on transition metal oxides as well as metal-organic complexes. There are optimal for solving several NP-hard problems especially in cyber security and bioinformatics (e.g. gene sequencing) where modern computing capabilities are falling short. Our existing measurement systems are not optimal for characterising metastable/ edge-of-chaos circuits. We are designing our own measurement circuits for characterizing such systems.

Reconfigurable computers: For handing big data in AI, attaining reconfigurability at a device functional level is crucial. In fact, despite many attempts, reconfigurability, parallelism and fault tolerance have been elusive goals in neuromorphic systems. We are designing materials with multiple thermodynamic transitions where logic parallelism and reconfigurability can be attained leading to substantial gain in computing energy, time and footprint.

Our expertise is primarily in designing devices and circuits. We look forward to industrial collaborations for interfacing with circuits, systems and software interfacing. We have adequate number of new and promising material platforms that could be translated to technologies.