Harnessing On-Device AI to Transform the Smartphone Experience
summary
Qualcomm shows on-device pipelines that move camera, audio and NLP from cloud to phone: a portrait-relighting flow combining demosaic → denoise → super-resolution → AI bokeh → relighting, and a low-latency speech-translation chain (ASR → NMT → TTS) running on accelerators. Claims: Snapdragon 888 brings a big jump (described as 26× performance/power step), tensor accelerator yields about 21× vs CPU, quantization covers >90% of networks, SNPE runs on 500M+ phones, and AIMET tooling and quantization flows are already used to ship 100+ real use cases.
AI-written summary · about
chapters (9)
- 0:00Introduction to on-device AI
- 4:05Trends driving on-device AI
- 13:00Key use cases and contributors
- 25:58Challenges of on-device AI
- 30:52Qualcomm's hardware solutions
- 33:59AI acceleration APIs and SDKs
- 37:51Platform solution ecosystem
- 42:24Summary and ecosystem impact
- 43:26Q&A session
related talks
16mRealize your vision with a guide to building Windows on Snapdragon
14mAccelerating On-Device AI Development
29mQualcomm® AI Stack for developers and extension to on-device AI
15mBuilding Gen-AI experiences using LLM and LVM
39mQualcomm Technologies and the power of Windows AI Q&A
08mUnlocking AI at Scale for PyTorch Developers
05mDrive UX innovation on Windows devices powered by Snapdragon
01h 20mWorkshop1: AI for Everyone