In this edition of the AI Technology Top Voice Interview series, we sit down with Ram Periakaruppan, Vice President and General Manager of the Wireline Network and Security business at Keysight Technologies. Ram has spent over two decades with Keysight (and previously Ixia), leading teams that develop cutting-edge solutions to help customers accelerate innovation cycles in network and data center technologies.
In our conversation, Ram shares how Keysight is shaping the future of AI infrastructure with its Keysight AI (KAI) architecture and flagship solutions like the KAI Data Center Builder and INPT-1600GE platform. We dive into how his team is helping customers validate performance, optimize configurations early in the design cycle, and accelerate the deployment of next-generation AI data centers. Ram also discusses the importance of full-stack validation, the role of automation in network testing, the security challenges of AI infrastructure, and his predictions for the future of AI-powered networks.
About Ram Periakaruppan: Ram Periakaruppan is the Vice President and General Manager of Keysight’s Wireline Network and Security business within the Communications Solutions Group. Based in Calabasas, California, Ram oversees R&D, product management, and business teams serving global industry verticals. With more than 24 years at Keysight and Ixia, he has held multiple leadership roles in engineering and product management, driving innovation in network test solutions that help customers bring first-to-market technologies to life.
About Keysight Technologies: Keysight Technologies (NYSE: KEYS) is a global innovation partner delivering solutions that accelerate innovation in communications, electronics, and AI infrastructure. With its end-to-end KAI architecture, Keysight enables organizations to validate, optimize, and secure next-generation data centers, networks, and applications. Its comprehensive portfolio spans compute, interconnect, network, and power domains, helping customers maximize performance, reduce risk, and speed time-to-market.
Here’s the full interview.
AI Technology Insights (AIT): Hi, Ram, Welcome to the AI Technology Top Voice Interview Series. Can you tell us about your current role at Keysight and what led you to this position?
Ram Periakaruppan: I am the Vice President and General Manager of our Wireline Network and Security business, which is part of Keysight’s Communications Solutions Group. I’m based in Calabasas, California and have been with the company for over 24 years, including my tenure at Ixia since 2001.
In my current role, I have full P&L responsibility for the business, overseeing R&D, product management, and business teams across key industry verticals. My teams are focused on delivering cutting-edge solutions that accelerate customer innovation cycles. Over the course of my career, I’ve held multiple leadership roles in engineering and product management.
Today, the wireline market is being reshaped by the demands of AI. I am spearheading Keysight’s efforts to evolve traditional emulation products to enable customers to bring first-to-market technologies to life — powering next-generation AI data centers and helping operators maximize the value of their infrastructure investments.
Recommended: AITech Top Voice: Interview with Rick Rosenburg, VP for Government Solutions at Rackspace
AIT: Keysight recently introduced the Keysight AI (KAI) architecture. Can you briefly explain what this is and where your team’s products fit into this architecture?
Ram Periakaruppan: To help optimize the AI infrastructure, we introduced our new architecture – Keysight AI (KAI) and its key foundational solution, which is KAI Data Center Builder. KAI is a suite of solutions designed to address the challenges of early validation, starting with pre-silicon, to packaging and system qualification.
The broad KAI architecture focuses on four areas – Compute, Interconnect, Network, and Power, each with its own powerful portfolio of solutions. The organization I lead offers solutions that fall under the interconnect and network pillars of the KAI architecture. I’ll discuss those products in more detail later. Keysight is the only company that delivers a full breadth of solutions across these four areas, covering every layer of the technology stack —from physical infrastructure to application performance.
With a long history in the network test space, we emulate anything and optimize everything across Layers 1 through 7 —helping our customers reduce workload risk and improve the efficiency of their AI enabled networks and data centers.
AIT: What are your team’s key offerings within the KAI architecture? How do these network validation solutions improve the overall product development lifecycle for your customers who build and operate AI data centers?
Ram Periakaruppan: The massive scale and complexity of AI infrastructure poses challenges that can make it difficult to train foundational models or serve thousands of customers in multi-tenant GPU clouds —especially under prolonged periods of maximum infrastructure utilization.
The industry has recognized that these highly interconnected systems demand full-stack validation and optimization. At the same time, to avoid costly delays and rework, there is a growing need to shift the validation process to earlier phases in the design and manufacturing cycle —well before all components are deployed at scale.
At Keysight, we are tackling these challenges with our KAI suite of products including my team’s KAI Data Center Builder – which enables users to design, validate, optimize, emulate, and benchmark real-world AI infrastructure at scale. Within the KAI portfolio, we also introduced our Interconnect and Network Performance Tester 1600GE (INPT-1600GE), a high-density, multi-port 1.6 Terabit Ethernet (1.6T) hardware test platform. The INPT-1600GE conducts traffic generation and analysis for functional performance and interoperability benchmarking across a broad range of silicon chips, interconnects, active cables, and networking equipment.
These KAI solutions help teams to identify bottlenecks, validate performance, and optimize their configurations early in product development. By shifting these critical steps forward, our customers reduce risk and streamline development while ensuring compliance at scale. This helps them avoid late-stage surprises and accelerate their time-to-market with greater confidence.
AIT: What role does the new INPT-1600GE platform play in accelerating the development and deployment of high-speed Ethernet AI data center and network infrastructures?
Ram Periakaruppan: The INPT-1600GE platform was purpose-built for the evolving demands of high-speed Ethernet validation and AI data center infrastructure. It is a portable, office-quiet solution that supports high-power optical receivers up to 40 watts, available in both portable benchtop and rackmount chassis. Whether it’s deployed in a lab or integrated into a larger test environment, the platform provides flexible port configurations—from 1x1600GE to 8x200GE—to support the broad range of devices using 212Gb/s electrical lane interfaces.
At speeds from 200GE to 1600GE PAM4, the INPT-1600GE assesses reliability, stability, and interoperability of silicon chips, active cables, optical transceivers, and networking equipment across Layers 1 through 3. The benchtop model offers greater mobility and ease of use. It includes a built-in handle, so engineers can move it around within the lab or outside the lab to validate components wherever it’s most needed.
The INPT-1600GE, combined with the Interconnect Test System (ITS) software that runs on the platform, enables critical interconnect performance evaluation and significant gains in testbed productivity. This equips our customers with the tools they need to deploy highly stable and reliable solutions into their networks.
Recommended: AITech Top Voice: Interview with Stephen Senterfit, President at Smartbridge
AIT: You mention the ITS software that operates on the INPT-1600GE platform. How does this software help improve efficiency and accuracy compared to more traditional validation methods? Does it differ from other solutions in the market?
Ram Periakaruppan: The Interconnect Test System (ITS) software transforms the INPT-1600GE platform into a powerful validation solution for high-speed Ethernet and AI data center environments. ITS is the first software of its kind to automate the process of collecting information about the interconnect, then uses that information intelligently to create test suites, or groups of tests that can run automatically which expedites test execution. Traditional approaches rely on manual scripting with limited traceability, but ITS offers an integrated system that intelligently organizes, stores, and leverages data—streamlining interconnect validation at scale.
ITS includes our U.S. patent-pending Interconnect Library (IL), which also improves productivity by allowing teams to reuse stored configuration and test results data for faster, more automated test case creation. The IL not only stores data, but it also remembers everything about each interconnect they measure. It strengthens traceability and provides full access to CMIS data and configuration history—all within an intuitive user-managed environment.
To increase throughput, engineers can queue and run scripts simultaneously across shared ports. Users can build pass/fail manufacturing test cases, generate automated reports, and even produce executable REST API scripts—without the need for programming expertise.
Together, the INPT-1600GE and ITS software deliver precise interconnect performance evaluations and significant testbed efficiency gains—helping customers deploy highly reliable solutions with greater speed and confidence. There are no other comparable solutions on the market right now.
AIT: What benefits does the KAI Data Center Builder solution deliver when it comes to emulating and validating real world AI workloads?
Ram Periakaruppan: To improve performance and reliability in AI data centers, testing must reflect real-world conditions as much as possible. Customers also need the flexibility to experiment with system design and workload optimization to achieve peak performance. That’s the core objective of the KAI Data Center Builder.
As the foundation of the KAI architecture, this solution emulates realistic AI workloads—including large language model (LLM) training jobs—without requiring massive GPU clusters. It enables users to evaluate how algorithms, components, and protocols impact system behavior, and to fine-tune infrastructure before full-scale deployment.
The KAI Data Center Builder provides the Collective Benchmarks application that measures and analyzes collective operations. This helps teams assess bandwidth consistency and job completion time across distributed AI workloads. With access to a library of workload traces and model partitioning schemas, users can experiment with traffic patterns, identify bottlenecks, and validate network fabric performance under high-scale conditions.
By bringing realistic AI workloads into the lab, the KAI Data Center Builder helps customers accelerate design cycles, reduce risk, and optimize infrastructure for scalable, efficient AI deployment.
AIT: What AI initiatives is Keysight working on in the U.S., and how will you help industries that are lagging in adoption?
Ram Periakaruppan: The industry is shifting to a multi standard ecosystem with new protocols. Some industry verticals not yet at the forefront of adopting AI technology are leveraging hyperscalers like Google GCP, Amazon, AWS, or Microsoft Azure to build out their AI inference capabilities. Cloud vendors like Lambda, CoreWeave or Crusoe also provide customized cloud AI environments for those who may not be able to build it out themselves. As businesses look for alternate technologies, we are also seeing things like InfiniBand, Broadcom’s new Scale Up Ethernet (SUE), NV Link from NVIDIA, and AMD’s UA Link, an alternate for scale up networking.
So, we have a very vibrant and rich environment of new networking protocols being developed to address the key networking challenges that AI presents. Keysight actively enables all of these technologies from physical layer solutions to conformance testing of the protocols as well as validating application performance and functionality.
There’s a whole ecosystem, both from an infrastructure standpoint as well as a system integration standpoint that they can tap into. As this ecosystem rapidly evolves, Keysight’s broad portfolio of network test and security solutions will ensure that the AI infrastructure is performing at its best to not only meet today’s needs, but to also scale for future requirements.
Recommended: AITech Top Voice: Interview with Paul Walker, Field CTO at Omada
AIT: Product teams are constantly concerned with security threats. How do you help keep AI infrastructure safe from the latest threats and risks?
Ram Periakaruppan: While the KAI Data Center Builder and INPT-1600GE focus on performance testing, we address AI infrastructure security with other Keysight solutions.
CyPerf, for instance emulates realistic LLM interactions and AI threats, including prompt-injection attacks, accidental data leakages, agent-to-agent (A2A) communications, and Model Context Protocol (MCP) workflow scenarios. This integrated approach enables organizations to measure the performance of their AI inference architecture while validating AI security gateways and related guardrails for their effectiveness in detecting both injections and leaks.
Backed by our Application and Threat Intelligence (ATI) team, with dedicated AI security researchers, CyPerf software continuously incorporates the latest injection techniques and LLM applications to ensure that defenses are tested against emerging threats, while also verifying their ability to handle legitimate traffic at scale.
AIT: What are your predictions for the future of AI-powered network technology?
Ram Periakaruppan: The future of AI-powered network technology will be defined by a multi-standard ecosystem designed to support both scale-up and scale-out networking for AI workloads. As AI models grow in size and complexity, the underlying network infrastructure must evolve to meet the demands of high bandwidth, low latency, and energy efficiency.
We’re already seeing the emergence of advanced technologies like silicon photonics and co-packaged optics, which enable faster and more efficient data transmission by integrating optical components directly with silicon chips. These innovations are critical for reducing power consumption and increasing throughput in data center environments. I anticipate that commercial deployment of these technologies will continue to grow over the next few years
I believe that 200G per lane electrical signalling will become a foundational building block for next-generation network interconnects. This paves the way for more widespread adoption of 1.6T Ethernet (1.6 terabit Ethernet) and 2 × 800GE (two 800 gigabit Ethernet links), which are essential for supporting the massive data flows required by AI training and inference clusters.
In parallel, we’ll see increased deployment of advanced optical technologies such as LPO (Linear Pluggable Optics) and LRO (Linear Receiver Optics). These solutions simplify optical transceiver design, which reduces power and latency—key benefits for hyperscale AI infrastructure. As the industry matures and ramps up production, these optics will become more prevalent over the next several quarters.
AIT: Can you recommend a few Keysight resources (whitepaper, case study, eBook, etc) that IT and AI leaders should read in 2025?
Ram Periakaruppan: Our team has created a lot of great resources so it’s hard to call out just a few, but I highly recommend these three papers:
- AI Fabric Test Methodology AppNote detailing specific KAI Data Center Builder test cases.
- Future Proofing Networks with 1.6T Ethernet whitepaper which dives into the challenges that INPT-1600GE and ITS software address in validating high-speed interconnects.
- High Performance Networking Offloads for AI / ML Focused Cloud Platforms AppNote.
This application note provides an in depth case study that shows how Keysight helped Crusoe identify bottlenecks between the public cloud and their data center infrastructure.
As an active member of the Ultra Ethernet Consortium (UEC), Keysight co-chairs various working groups within the UEC including the Performance and Debug and Physical Layer working groups. Each dynamic working group focuses on various aspects of Ethernet technology and helps shape the Ethernet landscape within everchanging AI and HPC environments. As a UEC member, we’ve published other documents about how to design and validate AI technologies. You can visit the UEC site for details on becoming a member or subscribing to their newsletter.
Finally, you can find additional resources or explore our full portfolio at: https://www.keysight.com/us/en/products/network-test.html
If you’d like to keep up with my team’s latest updates, you can subscribe to our monthly newsletter.
AIT: Tag a person in the industry whose answers you would like to see in the AI Tech Top Voice interview series:
Ram Periakaruppan: I would like to tag Dylan Patel, the founder of SemiAnalysis and Dr. J Metz, Senior Technical Director at AMD.
Thank you so much for your time today! We look forward to having you again at our Top Voice Series.
Recommended: AITech Top Voice: Interview with James Taylor, Head of Enterprise Transformation Platforms at DXC Technology
To share your insights, please write to us at sudipto@intentamplify.com
