Data & analytics at enterprise scale

I led a 30-person distributed engineering organization across Israel, France, Spain, and the US, delivering the analytics and anomaly-detection products that global enterprise customers ran their digital-experience decisions on.

Context

Behavioral analytics at this scale is a streaming problem: tens of millions of sessions a day, strict latency targets, and enterprise customers who notice every false alert. The anomaly engine was the product's signature feature and its biggest liability. It fired too often, took hours to surface anything, and trained customers to ignore it.

What I did

• Rebuilt anomaly detection. Replaced static thresholds with seasonal decomposition and per-metric baselines, then added a confirmation pass before anything alerted. False positives dropped 85%, and time-to-insight went from 12 hours to minutes.
• Modernized ingest. Led the migration of legacy ingest pipelines to a modern streaming architecture, improving data latency by 60% and laying the foundation for the next wave of products.
• Opened the data, not just the dashboards. Shipped Data Export integrations (Snowflake, BigQuery, Azure Blob) that let customers join experience data with their own warehouses. It became a key Enterprise-tier differentiator and influenced ~$12M in upsell and renewal revenue.

The leadership part

Scale is a people problem before it is a systems problem. I scaled teams through hypergrowth across seven countries, promoted eight engineers into senior and lead roles, and held voluntary attrition under 10%. Standards have to be the same in every time zone, and ownership has to be unambiguous when you cannot be in the room.

The proof I trust most came from the company's anonymous engagement survey, where every team was scored against the company baseline. As the outsider, not from the Paris HQ, running a team stitched together across a dozen nationalities, mine posted the strongest manager scores in the company: a perfect score on "my manager cares for my wellbeing as a person," and well above average on mentoring, support, and honest communication. The tell that the numbers were real and not a popularity halo: the same team rated company-level things, pay, the office, the strategy, bluntly low when they deserved it, and still rated their manager a 10. The outcome I am proudest of is not on the chart. People from other sites started asking to move onto the team, and you do not lobby to be managed by the outsider unless the standard inside the team is worth more than the geography.