Telemetry & analytics for real-time and AI-powered calling

Real-time calling produces an enormous volume of telemetry from two very different places: clients running on every platform, and the services behind them. Those sources disagree — different clocks, different sampling, different definitions of the same event.

Without a shared, trusted definition of reliability, latency and service health, teams end up arguing about whose dashboard is right instead of fixing the underlying product.

Led the design and implementation of telemetry, analytics and reporting solutions covering real-time communication and AI-powered calling scenarios.

Built Kusto-based telemetry pipelines that normalise and join client and service telemetry into consistent, queryable datasets.

Delivered Power BI dashboards on top of those pipelines so reliability, latency and service-health metrics are reviewable by engineers and leadership without writing a query.

Treated metric definitions as an interface: agreed semantics first, then implementation, so downstream consumers can rely on the numbers not shifting underneath them.

Engineering teams can measure reliability, latency and service health for real-time and AI-powered calling scenarios directly, rather than inferring them.

Telemetry became a first-class input to operational decision-making — deployment validation, alerting analysis and engineering health review.

Investigations into telemetry discrepancies now resolve to a root cause across sources instead of stopping at 'the dashboards disagree'.