AS
Back to Work

Outperforming Standard Observability

A root-cause engine that finds multi-variable failure paths that standard analytics platforms miss.

The problem

Standard streaming analytics platforms (like Conviva or Bitmovin) show that a metric dropped — but they fail to explain why.

When the issue is multi-variable (device × ISP × media × region × player), thresholds and dashboards don't isolate the true driver.

What I built / changed

  • Built a bespoke detection engine that treats root cause analysis as a supervised clustering problem.
  • Fit decision models to traffic data to identify "filtering paths" that deviate statistically from baseline.
  • Generated actionable slices like "a specific Media ID failing only on a specific User Agent".

Result

Autonomous isolation of complex, multi-variable issues that major industry tools miss entirely.

Stack / concepts

Supervised ClusteringPattern MiningAnomaly Detection

Interested in discussing engineering challenges?

Get in touch