WEKA is a cloud-native data platform company that delivers high-performance, software-defined storage and infrastructure purpose-built for artificial intelligence (AI), machine learning, and other next-generation compute workloads. Founded in 2013, the company began with a parallel file system and strategically transformed it into a scalable AI infrastructure platform. Today, that platform operates seamlessly across edge, core, hybrid, and multicloud environments. As a result, organizations can orchestrate low-latency, high-throughput data pipelines without operational friction.
At the core of WEKA’s innovation lies its NeuralMesh™ architecture. This architecture maximizes GPU utilization, eliminates data bottlenecks, and accelerates data-intensive workflows. Consequently, enterprises and research institutions can train, fine-tune, and deploy AI models faster and at scale. Leading global organizations rely on WEKA to power mission-critical AI initiatives, and the company continues to expand its international footprint with disciplined strategic growth.
In an exclusive conversation with The Interview World at the India AI Impact Expo 2026, Anthony Vandewerdt, Director of Solution Engineering at WEKA, articulated the company’s core technology vision. He detailed the differentiated capabilities that strengthen AI model development and production deployment. Furthermore, he shared a forward-looking perspective on the rapid evolution of India’s AI market. The following are the principal insights from that discussion.
Q: What are WEKA’s core technological capabilities and strategic focus within the technology domain?
A: WEKA delivers a software-defined storage platform engineered explicitly for artificial intelligence (AI), high-performance computing (HPC), and machine learning (ML) workloads. Unlike conventional storage systems retrofitted for modern compute demands, WEKA purpose-built its architecture for data-intensive, GPU-accelerated environments. As a result, the platform eliminates performance bottlenecks, sustains extreme throughput, and supports low-latency data access at scale. In essence, WEKA positions software, not hardware, as the defining layer of storage intelligence, thereby aligning infrastructure design directly with the performance economics of AI and HPC.
Q: What differentiated capabilities does your solution introduce that materially enhance AI model development, deployment, or operations?
A: Over the past decade, compute and network capabilities have advanced at an extraordinary pace. CPU and GPU performance has surged exponentially. At the same time, network bandwidth has scaled from 40 and 50 gigabits per second to 100, 200, 400, and now 800 gigabits per second. In parallel, PCIe technology has progressed from Gen 3 to Gen 4, Gen 5, and Gen 6, delivering unprecedented data transfer speeds within modern systems.
However, storage architecture has not evolved at the same velocity. Consequently, storage, not GPUs, not CPUs, and not the network, has emerged as the primary performance bottleneck in AI and ML environments.
Recognizing this structural imbalance, we concluded that incremental upgrades would not suffice. The industry required a fundamentally new storage architecture. Therefore, we engineered a platform purpose-built for the I/O patterns, concurrency demands, and throughput intensity characteristic of AI and machine learning workloads. Unlike legacy systems built on 10- to 20-year-old architectural assumptions, our design aligns natively with modern GPU-accelerated compute. As a result, it sustains performance at scale, eliminates data starvation, and keeps pace with the full potential of contemporary infrastructure.
Q: How does your solution enable AI models to scale sustainably while reducing cost and improving usability?
A: We operate as a core component of AI infrastructure. Specifically, we address the storage layer that underpins both training and inference workloads. In practice, organizations that provide GPU services or develop and deploy models encounter two dominant workload classes: training and inference. Each presents a distinct systems challenge.
In training environments, the primary objective is maximizing GPU utilization. Enterprises aim to drive GPUs as close to 100 percent utilization as possible. Otherwise, they compensate for inefficiencies by purchasing additional GPUs, an expensive and unsustainable approach. The constraint, more often than not, is not compute capacity but data delivery. When storage cannot feed GPUs at sufficient throughput and concurrency, expensive accelerators sit idle.
Inference introduces a different bottleneck. Here, memory availability and data access latency determine how quickly systems can return results. If the infrastructure cannot sustain high token throughput and low-latency reads, response times degrade. The consequence is operational friction and, ultimately, dissatisfied customers, either waiting hours for training runs and checkpointing cycles to complete or waiting too long for inference outputs.
We eliminate those constraints at the storage layer. In inference scenarios, our architecture enables response times measured in seconds and supports up to four times more token throughput compared to traditional storage infrastructures. In training environments, customers have reduced model epoch times from 80 hours to four by deploying our platform as the underlying storage foundation.
We do not replace GPUs, networking, or orchestration layers; those components remain essential. However, storage frequently represents the weakest link in modern AI stacks. By redesigning that layer for AI-native workload patterns, we remove the bottleneck, restore balance to the infrastructure, and unlock the full performance potential of the broader system.
Q: What is your outlook on the growth, maturity, and competitive landscape of the AI market in India?
A: The Indian market is expanding at an extraordinary pace. The scale and velocity of AI adoption are striking. Crucially, the Government of India has recognized artificial intelligence as a strategic force multiplier capable of accelerating economic growth and national competitiveness. As a result, it is committing substantial capital, policy support, and institutional focus to AI-driven initiatives. That commitment is evident across sectors—from public infrastructure and research to startups and large enterprises.
This momentum is precisely why we are in Delhi. We intend to participate in and contribute to this surge in AI-led transformation. The ecosystem’s energy is unmistakable. A broad spectrum of organizations has converged to invest, collaborate, and build. The level of engagement signals both urgency and ambition.
More importantly, AI’s potential impact on Indian society is profound. When deployed responsibly and at scale, it can enhance productivity, expand access to services, stimulate innovation, and elevate overall economic output. In turn, that progress can raise living standards and distribute opportunity more widely across the country. The trajectory is clear: India is not merely adopting AI; it is positioning itself to lead with it.
