Ola Mabadeje is a principal product management engineer at Outshift by Cisco developing AI systems for enterprise infrastructure and network operations. His work focuses on a consequential problem: understanding whether changes to complex production networks will cause failures before those changes are deployed.
Trained in electrical engineering, Mabadeje built his career in network engineering and Cisco product development; his professional background also includes Northwestern University’s Kellogg School of Management. He coauthored a 2013 analysis of mobile-network data analytics and wrote in 2017 about programmable networks and service-provider economics. At Cisco’s incubation group, Outshift, he subsequently helped develop Motific.ai, an enterprise generative-AI offering combining organizational data, security controls, and retrieval-augmented applications.
Building AI that understands infrastructure
- Network knowledge graphs grounded in real operations. Mabadeje’s architecture consolidates device configurations, network-controller data, security information, and streaming telemetry into an OpenConfig-based representation. The resulting network digital twin combines that graph with testing tools, letting agents examine relationships across network layers before proposed changes reach production. His team evaluated Neo4j and ArangoDB, initially selecting ArangoDB.
- Automated testing inside existing approval workflows. In his AI Engineer demonstration, a firewall-change request begins in ServiceNow. Specialized agents assess downstream impact, generate relevant tests, compare proposed configurations against a recent network snapshot, and return their findings to the original approval ticket. Human reviewers receive stronger operational evidence without abandoning established change-management processes.
- Cross-vendor agent interoperability. Mabadeje’s work connecting ServiceNow and Cisco agents addresses how independently developed enterprise agents can discover one another, communicate, and operate under shared identity and governance frameworks. He also found that fine-tuning a graph-querying agent on schema information and example queries reduced token consumption and response times.
In October 2025, Mabadeje coauthored a Swisscom–Outshift white paper on autonomous network operations describing a multi-agent network validator that uses knowledge graphs and digital twins to evaluate telecommunications infrastructure changes before deployment.
