This woman entrepreneur is building a many-core processor to enable businesses build energy-efficient, high-performance systems

With over 35 years of research and industry experience in computer architecture, parallel computing, dataflow computing, and fault-tolerant computing, Ranjani Narayan can be called a pioneer in the field of technology.

Born and raised in Bengaluru, Ranjani studied in the Kannada medium until high school and thought she would get an essential science degree from a local college because that was all the background she had. But fate, however, had significant plans.

She got admission into engineering at the Indian Institute of Science (IISc) in the late 70s and got her PhD degree in Computer Science and Automation from the same institute in the late 80s.

During her tenure at IISc, both as a student and staff, Ranjani worked on cutting-edge technologies, and later held essential stints at MIT, TU-Delft, and HP. In 2006, she started Morphing Machines with her husband, SK Nandy, Professor of Computation and Data Sciences at IISc.

Ranjani explains the pain point behind the venture.

“The broad topic of my PhD dissertation was computer architecture and the challenges associated with distributed computing. A key study that I did was find an inflexion point, where we get most in a distributed environment. This is a classic problem that needs a lot of attention. I have also introduced the then-emerging execution paradigm, the dataflow execution – this is the most natural way of executing a user application program,” she says.

Not surprisingly, issues identified with distributed computing remain the same even with technological improvisations over the years. Ranjani believes the challenges have only got more complicated. There is now an ever-growing need for highly available systems, support for emerging applications, need for reconfigurable processors, secure and safe computing - these have made a list longer.

Towards the end of her tenure with an MNC, she got an opportunity to conceptualise and build a future-proof processor through Morphing Machines. Incubated at IISc, its vision is to redefine high-performance computing. And thus, REDEFINE, the future-proof many-core processor, was born.

Building an all-core processor

Morphing Machines has recently partnered with SFAL (Semiconduction Fabless Accelerator Lab) – a Govt. of Karnataka and Indian Electronics & Semiconductor Association (IESA) initiative. It is all set to fabricate a Rs 256-core REDEFINE to meet the computing demands of emerging applications in AI/ML, IoT, and deeptech.

REDEFINE, indigenously conceived, developed, and implemented in India, is an excellent replacement and upgrade for CPUs and GPUs from multi-national brands.

Ranjani explains, “It is a multi-accelerator and many-core system-on-chip for compute-intensive applications on the Edge and at the Core. High-performance computing applications from the domain of AI/ML, and real-time reactive control applications for avionics and autonomous vehicles are candidates for acceleration on REDEFINE. REDEFINE can be configured into domain-specific accelerators (DSAs) on aggregates of computing cores through run-time specialisation. DSAs on REDEFINE seamlessly scale without any software changes. Multiple and diverse DSAs are realised on the REDEFINE SoC as software-defined domain-specific accelerators.”

She claims that REDEFINE-­based solutions offer a unique combination of high performance and flexibility not available from other technologies provided by global players. REDEFINE SoCs accelerate entire classes of compute-intensive applications at significantly low power.=

“Unlike other commercially available high-performance processors from multi-national brands, REDEFINE enables post-silicon customisation for strategic applications. REDEFINE by design is a secure computing platform architected for high availability and natively supports N-modular redundancy for safety-critical applications. REDEFINE is resilient to cyberattacks and naturally supports reactive applications using a data-driven model of execution,” she adds.

The journey, however, has been long and arduous. Many times, the founders thought of giving up and going back to their comfortable jobs.

“But it was our idea, our technology, our team and partners that kept us going. Our initial breakthrough has been our initial customers. Our recent association with SFAL has facilitated us to go further.

“Initially, we approached investors with our grand goal of going to the fab, which requires huge investments. Most of the Indian investors acknowledge openly that ours is a disruptive technology, and we can grow big and global. However, they seem to be risk-averse to the semiconductor business,” she says.

A reworked plan

This prompted Ranjani and Nandy to go back to the drawing board and revive their business plan. In the reworked plan, they have included several avenues of revenue generation and engaging with potential customers to have early customer lock-ins and early RoIs.

The founders are looking at REDEFINE increasing the global market. “This being an indigenous processor right from concept to implementation will be a proud moment for us, for the Silicon Valley of India, and mark India (in the global map) as one of the pioneers of reconfigurable many-core processors,” Ranjani says.

They are also part of Qualcomm’s QWEIN cohort for women entrepreneurs, with access to training and mentorship.

At this juncture, the founders are expecting two options to open:

Morphing Machines will continue to be an IP design and services company. It will continue to receive customer inputs, improve its product, and continue its educational services. It will also start to initiate the production of semiconductor SoCs based on REDEFINE.

In the next two to three years, they are looking forward to seeing a few soft and hard IPs of REDEFINE processors used by customers in real-life situations.

“We would hand-hold the customers for application development by providing an “Application Development Kit”. I would want to receive customer feedback, incorporate, and improvise the processor. One of the by-products during this phase will be an instrument for educational purposes. We will provide “educational kits” to universities to enhance knowledge of next-generation by offering a full-fledged semester-long course to cover all the important aspects of many-core processors – the challenges involved, the solutions that would address the challenges, etc.,” she adds.