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