Uploaded June 2026 | Updated November 2026, 1 day ago
In this episode, Norbert and Jakub from our AI team break down real-world examples from Apple, Google, Meta, and Snapchat. They explain why running AI locally delivers better privacy, faster performance, and offline capabilities, and show you how to start building these kinds of AI features yourself! π€
Follow us β¬οΈ
Software Mansion: https://x.com/swmansion
Kuba: https://x.com/_jakubchmura
Norbert: https://x.com/Nklockiewicz
Check out RN ExecuTorch: executorch.swmansion.com
Timestamps
00:00 Introduction to on-device AI and its misconceptions
00:47 What is on-device AI and its key benefits
01:25 Challenges of on-device AI: storage, battery, heat
02:17 Snapchat lenses: running models directly on device
03:28 Commercial use cases: virtual try-on and AR filters
04:54 Efficiency of on-device models for streaming and real-time processing
05:25 Semantic search in Apple Photos: how it works
06:29 Apple's approach to AI: hardware integration and privacy
07:21 The role of small LLMs and fine-tuning for privacy
08:43 Privacy and security benefits of on-device LLMs
09:49 Scam detection in messaging apps: real-time, on-device
11:07 Google Chrome's AI APIs for developers
12:20 Meta's on-device AI: optimizing infrastructure and user features
13:23 Instagram's object segmentation and stickers
14:24 Transcription of audio messages in messaging apps
15:10 Wearables with AI: AR glasses and live translation
16:35 The rapid evolution of AI tooling and APIs
17:24 Summary and future outlook of on-device AI
#ondeviceai #privacy #mobileai #semanticsearch #llms #aidevelopment #softwaredevelopment #pytorch #edgeai
In this episode, Norbert and Jakub from our AI team break down real-world examples from Apple, Google, Meta, and Snapchat. They explain why running AI locally delivers better privacy, faster performance, and offline capabilities, and show you how to start building these kinds of AI features yourself! π€
Follow us β¬οΈ
Software Mansion: https://x.com/swmansion
Kuba: https://x.com/_jakubchmura
Norbert: https://x.com/Nklockiewicz
Check out RN ExecuTorch: executorch.swmansion.com
Timestamps
00:00 Introduction to on-device AI and its misconceptions
00:47 What is on-device AI and its key benefits
01:25 Challenges of on-device AI: storage, battery, heat
02:17 Snapchat lenses: running models directly on device
03:28 Commercial use cases: virtual try-on and AR filters
04:54 Efficiency of on-device models for streaming and real-time processing
05:25 Semantic search in Apple Photos: how it works
06:29 Apple's approach to AI: hardware integration and privacy
07:21 The role of small LLMs and fine-tuning for privacy
08:43 Privacy and security benefits of on-device LLMs
09:49 Scam detection in messaging apps: real-time, on-device
11:07 Google Chrome's AI APIs for developers
12:20 Meta's on-device AI: optimizing infrastructure and user features
13:23 Instagram's object segmentation and stickers
14:24 Transcription of audio messages in messaging apps
15:10 Wearables with AI: AR glasses and live translation
16:35 The rapid evolution of AI tooling and APIs
17:24 Summary and future outlook of on-device AI
#ondeviceai #privacy #mobileai #semanticsearch #llms #aidevelopment #softwaredevelopment #pytorch #edgeai










