Case studies

The problems we found, the systems we shipped, and the numbers our clients measured.

Consumer app

Content platform, multi-million users

What we diagnosed: Pinpointed exactly where the recommender was losing users, across recall coverage, ranking, diversity, and cold-start.

+2%
retention lift
3-4 wks
to delivery
DeliveredRecsys diagnosis and tuning
Consumer app

Durable data layer and dashboards

What we diagnosed: Traced untrustworthy dashboards to upstream schema drift, then mapped the true cost of every query and the hours the whole company lost to it.

74%
lower data + compute cost
141 hrs
saved weekly across 20 people
DeliveredDaily aggregation, historical backfill, a new storage layer, and rebuilt dashboards
Gaming

AI ad-video generation

What we diagnosed: Broke down the cost structure of the video-generation pipeline and how often it had to retry to land a usable asset.

80%
lower video generation cost
Fewer
retries per usable asset
DeliveredEnd-to-end script to video to finished asset
Gaming

Ad-ops automation

What we diagnosed: Found the manual bottlenecks across the ad-ops pipeline and how many hours each one was quietly eating.

~120 hrs
saved weekly (about 3 FTE)
DeliveredEnd-to-end ad-ops automation