Prashant Mahajan

Content Identity Resolution

One program, one identity, across every screen.

Viewing data arrives with the same program titled in many different ways. This production system uses AI to resolve more than 10 million titles to a single canonical identity, so viewing on TV and on digital platforms is credited to the same program.

1. 01Raw titlesSame program, many names
2. 02AI matchingTrue duplicates grouped together
3. 03One identityCanonical ID across TV and digital

10M+

titles resolved

Production

live system, not a prototype

Hadoop to AWS Migration

Two years, 40+ Spark jobs, zero downstream disruption.

Migrated batch processing from an on-prem Hadoop cluster to AWS while product delivery continued. Downstream consumers saw no schema changes at cutover, and the legacy cluster was decommissioned.

40+

Spark jobs migrated

2 years

program duration

0

schema changes for consumers

Cross-Platform Campaign Results

Campaign measurement, from raw data to client reporting.

Comscore's cross-platform campaign measurement product. Owned the data pipeline end to end and later rebuilt it on AWS. Precomputing shared metrics once, rather than for each report, cut processing time roughly in half.

1. 01IngestAd and viewing data
2. 02ProcessDistributed Scala and Spark jobs
3. 03PrecomputeShared metrics, calculated once
4. 04ServeWarehouse and caching layer
5. 05ReportClient dashboards

~50%

less processing time

End to end

ingestion to reporting