Network Copilot™

Episode 2: Revolutionizing Networks with Network Copilot™: Insights from Chid Perumal - YouTube

Tap to unmute

Episode 2: Revolutionizing Networks with Network Copilot™: Insights from Chid Perumal Aviz Networks

Aviz Networks1.71K subscribers

Transcript

00:00:03

Ilona Gabinsky: Hello and welcome back to the Aviz Networks podcast series. I'm your host, Ilona Gabinsky, and today we're diving deep into our revolutionary GenAI-based Network Co-Pilot with none other than Chid Perumal, the CEO of Aviz. Chet, welcome to the show.

00:00:19

Chid Perumal: Thank you, Ilona. It's great to be here.

00:00:26

Ilona Gabinsky: Chet, as you know, AI is fundamentally changing the relationship between technology and humans—and for the first time, you can interact with technology like a coworker. What makes Aviz Network Co-Pilot so special in this evolving landscape?

00:00:57

Chid Perumal: In simple terms, Aviz Network Co-Pilot is an AI-powered assistant for your network. It’s smart and proactive, evolves with your network, and improves productivity by solving various use cases through a simple conversational chat interface. It’s like having a GenAI-skilled network engineer working alongside you 24/7.

00:01:21

Ilona Gabinsky: That sounds impressive. How can customers start integrating it into their existing infrastructure?

00:01:34

Chid Perumal: First, you need a truly open system—ready to adopt rapid innovations in the LLM space, including selecting the right model among many. Second, it must be enterprise-ready: a modern, scalable, secure, and robust architecture. Third, it should integrate easily with the existing environment so that introducing GenAI does not affect production networks. Finally, because it’s all about data, it must work across heterogeneous devices, operating systems, and tools.

00:02:40

Ilona Gabinsky: So, to start with Network Co-Pilot, you need a system that works with any LLM, is enterprise-ready, and integrates easily with your current infrastructure.

00:02:55

Ilona Gabinsky: I’m sure it can solve many problems for our customers. Can you mention a few?

00:03:02

Chid Perumal: Absolutely. Examples include: (1) network upgrade compliance checks—after software or hardware upgrades, repeatedly verifying the network’s operational behavior hasn’t changed; (2) security audits—continuous checks for configuration and operational compliance to avoid security violations; and (3) performance monitoring of ISPs—correlating data across routers, switches, multiple data sources, and tools in true multivendor environments.

00:04:24

Ilona Gabinsky: People also ask about prompt engineering. How does prompt engineering work with Network Co-Pilot?

00:04:49

Chid Perumal: Prompt engineering is critical and follows a well-defined process. First, clearly understand the problem to solve. Second, identify the datasets and where to collect them. With lots of data, you must ensure accurate responses—create the right SQL queries and reduce hallucinations via model tuning. At Aviz, we use techniques such as few-shot examples and ReAct prompting for accurate SQL generation, and a RAG approach (retrieval-augmented generation) to ground responses with the right context.

00:06:06

Ilona Gabinsky: So it’s like asking ChatGPT a question—it won’t know the answer unless it has the data and has been trained or fine-tuned on it.

00:06:13

Chid Perumal: Correct. It’s all about high-quality data for fine-tuning and grounding.

00:06:21

Ilona Gabinsky: For customers who already have many network tools, how can Network Co-Pilot integrate with them?

00:06:30

Chid Perumal: Network Co-Pilot is powered by a modern agent framework that’s becoming popular in the GenAI space. With this approach, existing automation and NetOps tools can be integrated easily. At Aviz, we’ve integrated our own tooling so Co-Pilot can automatically create orchestration templates over a chat interface.

00:07:04

Ilona Gabinsky: So we can make existing tools essentially powered by LLMs.

00:07:11

Chid Perumal: Yes—that’s where the future is heading.

00:07:16

Ilona Gabinsky: I just came back from Cisco Live, where customers were impressed with our Network Co-Pilot demo. A frequent question is data security. How do we ensure customer data is safe?

00:07:49

Chid Perumal: Data security is a key tenet of Network Co-Pilot. It ships as a standalone package for on-premises deployment without requiring internet access. We will never call an outbound API with customer data—data stays on-prem, under the customer’s control. We also provide enterprise-grade security features such as RBAC, secure APIs to access the datastore and interfaces, and other hardening measures.

00:08:27

Ilona Gabinsky: That’s very important for our customers. Another question: how do we connect to the data sources?

00:08:48

Chid Perumal: Connecting to data sources is straightforward, and collection must not disrupt the existing infrastructure. Network Co-Pilot provides data collection services that integrate with standard mechanisms like SNMP, APIs, gNMI, and flow records. Our ingestion pipeline also accommodates customer-specific tools.

00:09:23

Ilona Gabinsky: Here’s a specific use case: customers want to know if their firewall is overloaded. Can we help?

00:09:40

Chid Perumal: Yes. Co-Pilot can determine that by analyzing network traffic data received by the firewall, system resource utilization (CPU, memory, disk), relevant hardware tables, and by applying a clear definition of what “overloaded” means in that environment.

00:10:13

Ilona Gabinsky: So we can use GenAI to solve many problems based on customer requirements, and they can build their own use cases.

00:10:25

Ilona Gabinsky: One more: how does Network Co-Pilot learn—does it learn continuously?

00:10:33

Chid Perumal: Yes. Network Co-Pilot uses continuous learning to adapt based on incoming (near-)real-time data. We also improve LLM responses by enforcing guardrails to follow interface-specific, legal, and ethical rules, maintaining accuracy while the system learns.

00:11:02

Ilona Gabinsky: Last question: will AI replace the engineer?

00:11:16

Chid Perumal: AI is a tool to augment—not replace—human capabilities. Network Co-Pilot handles repetitive, routine, time-consuming tasks so engineers can focus on strategic innovation. It enhances productivity and empowers teams to tackle complex challenges more effectively.

00:11:38

Ilona Gabinsky: I couldn’t agree more. AI makes me more efficient, but it can’t replace the creativity and innovation humans bring.

00:11:54

Ilona Gabinsky: Before we wrap up, any final thoughts for our listeners?

00:12:03

Chid Perumal: Network Co-Pilot is a significant leap forward in how we manage and optimize networks. By leveraging GenAI, we help businesses achieve greater efficiency, reliability, and security in network operations.

00:12:18

Ilona Gabinsky: Chet, thank you for your time and for sharing your expertise. And thanks to our listeners for tuning in. Stay tuned for our next episode, where we’ll explore more about AI Fabric and data-driven network applications—and much more. Thank you!