Watch on YouTube
Watch on Vimeo
In this session, Dr. Deepak Kakadia, founder and CEO of NetAI, discusses the technical architecture of NetAI’s graph neural network (GNN) and how it provides deterministic root cause analysis for autonomous network operations. Kakadia leverages his experience at Sun Microsystems, Verizon Labs, and Google to explain why traditional AIOps and Large Language Models (LLMs) often fail in networking environments. He argues that while LLMs are designed to model human language and behavior, networks are human-built structures that are better represented as mathematical graphs, allowing for a more precise and deterministic approach to troubleshooting.
The presentation details how NetAI’s GNN-based engine captures the structural relationships between routers and edges, mapping protocol layers and topology to identify exact root causes rather than statistical guesses. Unlike LLMs, which require exhaustive training on every possible permutation of alarms and symptoms, GNNs utilize the inherent causal relationships of the network to provide verifiable diagnostics. This approach eliminates the “best guess” nature of probabilistic models, reducing the burden on network engineers who would otherwise have to manually verify AI-generated suggestions. The system acts as a digital twin that records the state of the network at every timestamp, enabling historical analysis of intermittent issues that are notoriously difficult to replicate.
Kakadia emphasizes that NetAI is a product-focused company offering a rapid-deployment, containerized solution that can run on-premise in air-gapped environments or in the cloud. By integrating with existing observability data and automation scripts, NetAI fills the gap between identifying a problem and taking corrective action, effectively enabling self-healing network capabilities. The session concludes by highlighting the tool’s ability to lower Mean Time to Repair (MTTR) and improve productivity by allowing engineers to focus on root causes rather than downstream correlated alarms. While focused strictly on the networking stack from Layer 1 to Layer 4, the platform provides deep insights that help organizations rule out network issues during complex application outages.
Personnel: Deepak Kakadia
Thank you for being part of the Tech Field Day community! Our mailing list is a great way to stay up to date on our events and technical content, and we appreciate your signup.
We promise that we’ll never spam you, send ads, or sell your information. This list will only be used to communicate with our community about our events and content. And we’ll limit it to no more than one message per week.
Although we only need your email address, it would be nice if you provided a little more information to help us get to know you better!
