PocketAI Hub
Private, on-device AI for chatting, creating media, and running agents on the hardware you already own.
- Platforms
- 5
- Compute
- On-device
- Role
- Creator
Overview
What it is
PocketAI Hub brings private AI to iOS, Android, macOS, Windows, and Linux. Models run on the user's own hardware, work offline after download, and support chat, media creation, agents, and secure access to a more powerful desktop.
The interesting part
The challenge
Local AI has to respect the limits of each device without making the product feel limited. That means matching models to hardware, packaging runtimes cleanly, and making privacy, downloads, tools, and multimodal work understandable to normal users.
Selected work
What I built
- Built the cross-platform product experience around local model discovery, downloads, inference, conversations, media generation, and agent workflows.
- Created the supporting systems for remote desktop access, device-aware model choices, offline operation, and sensitive-action controls.
- Converted, quantized, packaged, benchmarked, and validated MLX and GGUF model releases, including Apple-Silicon-native video generation models published on Hugging Face.
Result
Outcome
PocketAI Hub is live across five operating systems and is growing into a practical private-AI platform. The public Hugging Face work documents the model engineering behind it, not just the interface around it.