Aviz Office Hours Recap: Exploring Network Copilot™ and AI-Powered Networking
January 5, 2026
Network Copilot ™
Summary
Aviz NCP Office Hours held in December (APAC) was a virtual interactive session designed to ask, learn, and experiment with Network Copilot™. The discussion focused on AI-powered networking, multi-vendor intelligence, and a behind-the-scenes look at how Network Copilot™ works under the hood, delivered through an open live Q&A format and a short real-world demo.
Introduction
Aviz NCP Office Hours provided a dedicated space for open discussion around Network Copilot™. The session emphasized direct interaction over formal presentations, allowing participants to engage with the Aviz team through live questions and real-time discussion.
The event was conducted online and followed a conversational format, with no slides and no recordings, enabling focused participation and candid exchanges.
Speakers
Madhu Paluru
Director of Engineering, Aviz Networks
Madhu Paluru is a technology leader with more than 20 years of experience across Cloud, Data Networking, and Open Source. At Aviz, he drives a GenAI-powered Network 3.0 stack and moderates within the SONiC community.
He has trained over 100 engineers and helps global teams operationalize SONiC, with a strong focus on innovation and community building.
Cody McCain
Director of Product Management, Aviz Networks
Cody McCain is a physicist turned networking leader with more than 20 years of experience across networking, cybersecurity, and software engineering. He has held roles at JPMorgan Chase, Intel, Cisco, VMware, and Tigera.
His experience includes protecting $11 trillion in daily network transactions, leading CNI projects such as Cisco Network Operator and Antrea, and working on Kubernetes and container security. At Aviz, he leads AI-driven solutions focused on simplifying networking operations and infrastructure automation.
What Was Covered During Aviz Office Hours
The discussion was driven by live questions and practical topics related to Network Copilot™.
Key Discussion Areas
- How to deploy and bring up Network Copilot™ on-site
- Using roles, permissions, and projects to run tasks in parallel
- How data connectors work, including attributes, normalization, MCP, and API approaches
- Leveraging file uploads for context using RAG
- Agents in Network Copilot™, including available options, building custom agents, and SDK walkthrough
- LLM options in NCP and their pros and cons
- Prompting techniques that work
Live Demo and Interaction Format
Session Format
The Office Hours followed a simple and interactive structure:
- Live Q&A throughout the session
- A 10-minute demo walking through a real-world use case
- No slides and no recordings
- A session summary shared after the event
This format allowed participants to focus on questions, discussion, and live interaction during the session.
Conclusion
Aviz NCP Office Hours – December (APAC) delivered an open forum for discussion, learning, and experimentation with Network Copilot™. Through live Q&A and a short demo, the session provided clarity on deployment, data connectors, agents, and LLM usage within Network Copilot™.
The event reinforced Aviz Office Hours as a space for direct engagement and practical exploration of AI-powered networking concepts.
Frequently Asked Questions
Why do traditional network operations tools struggle to keep up as environments grow more complex?
Engineers operate in environments filled with dashboards, logs, alerts, and tools that each show only part of the picture. Constant context switching between those tools reduces accuracy, increases mistakes, and leads to burnout. The data is all there but it is fragmented across systems that were not designed to talk to each other, so engineers end up doing the correlation work manually rather than having it done for them.
How does AI actually help network engineers rather than replacing them?
AI augments what engineers can do rather than substituting for their expertise. It synthesizes growing telemetry volumes, adds context to alerts, and helps teams reason through issues faster. The episode is direct that AI is not optional for NetOps anymore but it is also clear that the goal is smarter decision-making by engineers, not removing engineers from the decisions.
What tasks are the best fit for an AI co-pilot in network operations?
Chid calls out three specifically: compliance checks, anomaly summarization, and alert correlation. These are all tasks that require pulling together data from multiple sources and applying consistent logic to it. They are time-consuming when done manually and prone to inconsistency when different engineers interpret the same signals differently. AI handles that synthesis reliably at a pace no human team can match.
Why do different teams interpret the same network data differently and what problems does that cause?
Network, security, and operations teams all look at the same signals through different lenses based on their priorities and experience. A metric that looks normal to one team is a red flag to another. That inconsistency slows down incident response and makes it harder to reach shared conclusions. AI adds context and explanation to the data so different teams are working from the same understanding rather than arguing about what a signal means.
Why does an effective AI co-pilot need to be vendor-agnostic rather than tied to one platform?
A vendor-specific AI co-pilot just creates another operational silo. If the AI only understands one vendor's data and tools, teams are back to manual correlation for everything outside that vendor's environment. Vendor-agnostic AI works across the whole environment so engineers get a unified view rather than a new fragmentation problem sitting on top of the old one.
Why is constant context switching between tools such a serious problem for network engineers today?
Every time an engineer switches between tools to piece together what is happening, they lose context and introduce the possibility of missing something. At scale with hundreds of alerts and multiple dashboards, that cognitive load compounds quickly. Accuracy drops, mistakes happen, and the sustained pressure leads to burnout. The episode frames this as one of the core reasons traditional operations no longer scales as environments grow.
Why are clean and scalable data pipelines described as the foundation of AI adoption in NetOps?
AI is only as useful as the data it reasons over. If the telemetry pipeline is messy, incomplete, or inconsistent, the AI co-pilot produces unreliable outputs and engineers stop trusting it. Clean and scalable data pipelines are what make AI adoption practical rather than a proof of concept that works in demos but fails in production where the data is noisy and the environment keeps changing.
What does model flexibility mean for NetOps teams choosing an AI co-pilot?
Model flexibility means teams can choose AI models based on their own cost, privacy, and performance requirements rather than being locked into one model chosen by the vendor. A team handling sensitive network data may need a private on-premise model. Another team may prioritise inference speed over cost. Flexibility lets the AI co-pilot adapt to what the environment actually requires rather than the environment adapting to the AI.
Missed the December (APAC) session?
Join the upcoming Network Copilot™ Office Hours in January to ask questions, explore new use cases, and see Network Copilot™ in action.
Register now
