Vinod Kannan
VP & GM - Compute BU
In 2009, when multicore computing became mainstream, Multicoreware began its journey. The founders identified a gap between rapidly evolving hardware and software that was struggling to harness it. As processors diversified into CPUs, GPUs, DSPs, and NPUs, performance was no longer dictated by raw hardware but by how intelligently software could align with each core’s unique characteristics. From that realization grew an organization focused on one principle: make software fit the silicon, not the other way around.
Vish Rajalingam, VP & GM - Mobility &Transportation BU
Engineering for Heterogeneity
Multicoreware is a global software IP and engineering services company with development teams spread across five centers worldwide. Its portfolio spans AI computing for edge and cloud infrastructure , mobility and transportation, media and entertainment, and Industry 4.0 applications. The company’s customers include semiconductor leaders, along with deep-tech AI startups and automotive OEMs such as AMD, Renesas Electronics, ECarX, Cadence and Lumotive and more.
Deep cross-functional expertise in signal processing, image processing, and computer vision is the discipline that enables performance optimization for real-world AI deployments at the company. These technical foundations position Multicoreware as an embedded partner that helps design and tune AI pipelines for production-grade efficiency across heterogeneous silicon.
Its teams specialize in AI model optimization, using techniques such as quantization, pruning, and compiler-level tuning to maximize throughput and reduce latency. Each customer engagement includes what the company calls a Model Zoo, a curated ecosystem of pre-optimized AI models tailored to a specific hardware platform. This approach transforms hardware constraints into design opportunities, ensuring that AI workloads run efficiently, predictably, and at scale.
The Cloud-to-Car Continuum
Multicoreware’s understanding of SDV architecture extends beyond the vehicle itself. The company has been shaping what it terms the Cloud + Embedded Compute continuum, a framework that enables seamless integration of automotive intelligence, allowing it to be built, simulated, and deployed between the cloud and the car.
Its engineers design sovereign or local cloud infrastructures for AI workloads, an increasingly critical need as data privacy and intellectual-property protection become strategic priorities for automakers. The company explains that its frameworks enable developers to run large-scale inference workloads, such as Level 3 + ADAS systems, through hardware-in-the-loop setups that replicate millions of virtual miles before physical testing begins.
Compute hardware doesn’t just play a role; it is the foundation upon which SDVs must be built”, says Vish Rajalingam, VP & GM - Mobility & Transportation BU, Multicoreware. “If the software pipeline is not architected from the ground up to exploit heterogeneous cores end-to-end, robust AI features are simply not possible”, Vish adds. This philosophy of cloud-validated, vehicle-optimized design has become central to how the company defines SDV readiness.
In practice, it allows automakers to simulate, test, and refine AI models at cloud scale, then port the verified software into the vehicle runtime with minimal friction. By bridging these environments, Multicoreware converts the SDV from an aspirational concept into an operational system that evolves continuously.
Open Innovation as Strategy
For Multicoreware, innovation operates on two fronts: intellectual property and open collaboration. The company’s contributions to open-source projects such as x264, x265, FFmpeg, HandBrake, OpenCL, LLVM, and MLIR are well documented and publicly available, reflecting a long-standing commitment to collective progress in computing performance. The company reports that work on the forthcoming x266 codec continues this tradition of open, standards-based development.
Beyond media technologies, the company invests in proprietary AI IPs that extend into human-sensing and environmental intelligence. One example is its radar-based Occupancy, Presence, and Vital Signs Detection technology, designed initially for intelligent vehicle cabins but adaptable to robotics, smart-building, and industrial safety systems. The company states that these solutions are being developed in alignment with emerging regulatory frameworks on privacy and in-vehicle monitoring, an acknowledgment of the ethical landscape surrounding AI perception systems.
Multicoreware also participates as a Premium Member of the Autoware Foundation, contributing to software and systems working groups that advance high-performance AI stacks for intelligent vehicles. “This participation reinforces a core belief that meaningful AI impact comes from collaboration, hardware-software co-design, and execution efficiency”, says Vinod Kannan, VP and GM - Compute BU at Multicoreware.
Every compiler refinement and model-optimization step serves the goal of maximizing performance per watt, per dollar, and per design constraint. By narrowing the gap between ambition and implementation, the company turns the art of software optimization into a strategic lever for innovation, ensuring that the growing computational power embedded in modern systems serves a measurable purpose.
The Cloud-to-Car Continuum
Multicoreware’s understanding of SDV architecture extends beyond the vehicle itself. The company has been shaping what it terms the Cloud + Embedded Compute continuum, a framework that enables seamless integration of automotive intelligence, allowing it to be built, simulated, and deployed between the cloud and the car.
Its engineers design sovereign or local cloud infrastructures for AI workloads, an increasingly critical need as data privacy and intellectual-property protection become strategic priorities for automakers. The company explains that its frameworks enable developers to run large-scale inference workloads, such as Level 3 + ADAS systems, through hardware-in-the-loop setups that replicate millions of virtual miles before physical testing begins.
Compute hardware doesn’t just play a role; it is the foundation upon which SDVs must be built”, says Vish Rajalingam, VP & GM - Mobility & Transportation BU, Multicoreware. “If the software pipeline is not architected from the ground up to exploit heterogeneous cores end-to-end, robust AI features are simply not possible”, Vish adds. This philosophy of cloud-validated, vehicle-optimized design has become central to how the company defines SDV readiness.
In practice, it allows automakers to simulate, test, and refine AI models at cloud scale, then port the verified software into the vehicle runtime with minimal friction. By bridging these environments, Multicoreware converts the SDV from an aspirational concept into an operational system that evolves continuously.
Open Innovation as Strategy
For Multicoreware, innovation operates on two fronts: intellectual property and open collaboration. The company’s contributions to open-source projects such as x264, x265, FFmpeg, HandBrake, OpenCL, LLVM, and MLIR are well documented and publicly available, reflecting a long-standing commitment to collective progress in computing performance. The company reports that work on the forthcoming x266 codec continues this tradition of open, standards-based development.
Beyond media technologies, the company invests in proprietary AI IPs that extend into human-sensing and environmental intelligence. One example is its radar-based Occupancy, Presence, and Vital Signs Detection technology, designed initially for intelligent vehicle cabins but adaptable to robotics, smart-building, and industrial safety systems. The company states that these solutions are being developed in alignment with emerging regulatory frameworks on privacy and in-vehicle monitoring, an acknowledgment of the ethical landscape surrounding AI perception systems.
Every compiler refinement and model-optimization step serves the goal of maximizing performance per watt, per dollar, and per design constraint
Multicoreware also participates as a Premium Member of the Autoware Foundation, contributing to software and systems working groups that advance high-performance AI stacks for intelligent vehicles. “This participation reinforces a core belief that meaningful AI impact comes from collaboration, hardware-software co-design, and execution efficiency”, says Vinod Kannan, VP and GM - Compute BU at Multicoreware.
Every compiler refinement and model-optimization step serves the goal of maximizing performance per watt, per dollar, and per design constraint. By narrowing the gap between ambition and implementation, the company turns the art of software optimization into a strategic lever for innovation, ensuring that the growing computational power embedded in modern systems serves a measurable purpose.
