Anurag Dhingra

AI Summary

→ WHAT IT COVERS Anurag Dhingra explains Cisco's approach to integrating AI across enterprise networking and collaboration tools, including custom silicon development, domain-specific models for network management, and autonomous agents that simplify IT operations while maintaining human oversight. → KEY INSIGHTS - **Ultra-Reliable Wireless Backhaul:** Cisco developed technology beyond traditional Wi-Fi to provide uninterrupted connectivity for factory robots, preventing assembly line shutdowns where even brief network interruptions cause measurable production losses and require seamless roaming across manufacturing facilities. - **Domain-Specific AI Models:** Cisco built three specialized models—security (open-weight on Hugging Face), Deep Network Model for networking tasks, and Machine Data Model for Splunk observability—distilled from larger models and trained on 40 years of networking telemetry data. - **Programmable Network Silicon:** Custom chips in Cisco switches enable software upgrades to unlock new AI capabilities without hardware replacement, addressing the challenge that infrastructure investments last five to seven years while AI state-of-the-art evolves every six months. - **Agent-to-Agent Communication Standard:** Cisco contributes to the A-to-A standard enabling collaboration between specialized agents from different vendors, allowing network management agents to automatically update ServiceNow tickets after resolving issues, mirroring human cross-functional workflows. → NOTABLE MOMENT Dhingra reveals that video background blurring runs on a specialized AI model inside the camera itself, requiring minimal compute power because it only recognizes human silhouettes versus backgrounds—demonstrating how purpose-built models outperform general models for narrow tasks. 💼 SPONSORS None detected 🏷️ Enterprise AI Infrastructure, Agentic AI, Edge Computing, Network Security