Case 01
Product intelligence · Enterprise AI transformation
Make the expertise a system, not a person: Operational intelligence and agentic workflows.
The mess
Connected charging estates generate dense streams of fragmented telemetry: intermittent OCPP communication drops, phase imbalances, payment timeouts, and vehicle-side handshake terminations. Frontline engineering teams drown in repetitive investigations, relying on tribal intuition to distinguish transient network noise from catastrophic inverter hardware failure.
The architecture
Architected an operational intelligence platform converting tacit engineering diagnosis into repeatable software behaviour. Designed rules engines evaluating recurring charging-session heuristics, alert suppression hierarchies, and telemetry correlation models. Established the strategic roadmap progressing from deterministic threshold rules to predictive fault clustering, vendor-neutral SaaS ingestion, and governed agentic automation.
The trade-off
Deterministic explainability versus opaque machine learning velocity. Prioritised human-in-the-loop, explainable diagnostic evidence for operators and customer executives over black-box predictions, ensuring every automated escalation carries verified provenance.
Enterprise-funded operational intelligence initiative. Metrics and workflows describe telemetry analysis and software capability; commercial figures and partner identities withheld for confidentiality.