In Bengaluru, a fast-growing deep-tech startup is helping solve one of the biggest problems facing artificial intelligence today. Founded in Silicon Valley in 2019, d-Matrix is one of the industry’s fastest growing, highly regarded startups due to its fundamental re-design of how AI systems are architected, focusing exclusively on inference at scale. The company is led by industry veterans Sid Sheth and Sudeep Bhoja, accomplished technologists of Indian origin who saw early that GPUs alone would not be enough to support the next generation of highly interactive AI applications. That vision is now coming to life through Corsair, d-Matrix’s inference accelerator built on a patented memory-forward, in-memory compute architecture, and JetStream, the first network accelerator purpose-built for inference. We spoke with Pradip Thaker, Head of India at d-Matrix, about the technology, the opportunity, and what lies ahead in AI and India’s role in shaping it.

What problem is d-Matrix fundamentally trying to solve?

As enterprises race to deploy AI – most assume the challenge is compute. However, the real constraint is data movement. Enormous amount of data must move between memory and processors every time an AI model runs. That creates latency, is extremely power hungry, and limits scale. We call this the memory wall, and it’s now well known to be the primary bottleneck for AI inference. d-Matrix fundamentally addresses this challenge with In Memory Compute architecture that enables delivery of performance, power, cost and efficiency all at once for AI infrastructure, especially for inference.

What is Corsair, and who is it designed for?

Corsair is our inference accelerator, built specifically for companies running large-scale AI in production. It’s designed for cloud service providers, hyperscalers, AI platform companies, and enterprises that need to serve large volume inference requests efficiently. These companies care deeply about cost per query, response time, throughput, and energy efficiency—and that’s exactly what Corsair is built to optimize.

What makes Corsair different from traditional accelerators?

Corsair is built on a memory-integrated compute architecture that brings data and compute together. By reducing data movement, we achieve lower latency, higher throughput, and significantly better power efficiency. It’s a fundamentally different way of thinking about AI infrastructure.

You also mention JetStream. What is that, and why is it important?

JetStream is our network accelerator that sits alongside Corsair to optimize how data moves across the system. As AI models scale, the bottleneck is no longer just inside the chip—it’s across the entire rack and data center. JetStream is the first NIC purpose-built for inference workloads, helping move data between nodes faster and more efficiently. Together, Corsair and JetStream form a tightly integrated platform that removes both the compute and networking bottlenecks that limit AI systems today.

How do Corsair and JetStream work together in a real deployment?

Together, Corsair and JetStream dramatically reduce data movement across the chip, server, and rack– enabling faster, more efficient, and more scalable AI systems. It enables companies to do AI inference at record speed.

Why is this especially important now?

Training gets the headlines, but inference is where AI delivers value. As usage grows, costs grow exponentially unless you rethink the system architecture. That’s why inference is now the real battleground—and why architectural innovation matters more than incremental performance gains.

What role does India play in building this technology?

Our Bengaluru team is a core part of d-Matrix’s innovation engine. Engineers here work on hardware, software, and system architecture. This is not a support function—it’s foundational R&D. As we grow, India is becoming central to how we design and deliver next-generation AI infrastructure.

What does this mean for India’s position in the global AI ecosystem?

Pradip: India is moving into a phase of deep-tech leadership. The talent here is designing core technologies that will be deployed globally. That shift—from implementation to innovation—is incredibly important for the country and for the global AI ecosystem.

Also Read | BHASHINI: Bridging 1.4 Billion Citizens to Services and Opportunities via Language AI

What kind of people thrive at d-Matrix?

People who like solving hard, meaningful problems. This is system-level work, not incremental optimization. If you want to help build the infrastructure that powers AI worldwide, this is the place to do it.

What excites you most about what’s ahead?

The future of AI won’t be defined by bigger models alone. It will be defined by smarter, more efficient systems. From Silicon Valley to Bengaluru, we’re building that foundation—and we’re just getting started. d-Matrix’s current product lineup and future roadmap are dependent on significant contributions from the company’s rapidly growing engineering team in Bengaluru, which has become a core innovation hub. And d-Matrix is preparing to more than double its presence in India over the next year.

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter, Instagram.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!