Case study · Qualcomm
Snapdragon Moments
Frazil helped turn Qualcomm's on-device highlight detection engine into a shippable Android product for Snapdragon devices.
Client
Qualcomm
Platform
Android · Snapdragon
Privacy
Fully on-device
Delivery
Productized ML engine
The brief
Qualcomm had an engine that could analyze gameplay in real time and detect highlight moments. Frazil built the product layer around that capability: a polished Android surface, capture and playback flows, telemetry, and release paths for Snapdragon devices.
The core constraint was privacy and latency: the experience had to stay fully on-device, with no server round trip or cloud clip processing.
What Frazil built
From the engine boundary out to release readiness.
01
Productized the engine
Turned Qualcomm’s highlight-detection engine into an Android experience people could open, understand, and use.
02
Kept the core boundary clean
Wrapped the existing detection logic behind a stable integration layer so the product could evolve without changing the engine.
03
Prepared it for release paths
Built the capture, playback, telemetry, and distribution paths needed for a production mobile surface.
Outcome
Fully on-device highlight detection
Consumer app surface around an existing engine
One product architecture for multiple placement paths
On-device inference changes the product around it: permissions, storage, battery, sharing, error recovery, and release paths all have to assume the network is gone.