Inspiration
OmniMind was inspired by the need for a truly private, offline, and high-performance AI assistant that safely processes user data locally. By combining this with a customized physical robot head (InMoov framework with fully operational movable eyes and eyelids), we wanted to bridge the gap between abstract software models and engaging physical AI interactions, completely free from the constraints and privacy concerns of cloud computing.
What it does
OmniMind acts as an intelligent offline personal companion running on local hardware. It processes complex tasks and manages conversations seamlessly without requiring an internet connection. At the same time, it drives the physical movements of a robot head to reflect responsive, real-time gestures, providing an interactive, embodied AI experience that can assist users with maximum privacy.
How we built it
The software application is entirely engineered using the Flutter framework and Dart for a highly responsive user interface across desktop environments, managed with Git and developed inside VS Code. The core AI model is highly optimized for local inference on PC/Laptop hardware. On the hardware side, we utilize custom-engineered STL files to 3D-print and manufacture the mechanical structures of the InMoov robot head, coordinating specialized servo motors to sync with the AI's internal processes.
Challenges we overcame
One of the main challenges was ensuring the offline model runs smoothly with low latency on local CPU/GPU resources while keeping a small disk footprint. We spent significant time optimizing inference performance. Additionally, wiring and accurately mapping the internal hardware interfaces—ensuring stable synchronization between the software commands and physical actions—required precision engineering.
What's next for OmniMind
Our next immediate step is to deploy OmniMind onto specialized Arm-powered edge hardware like a Raspberry Pi or single-board computers to increase energy efficiency. We also plan to integrate the mechanical eyes and eyelids into broader adaptive physical behaviors, making OmniMind the ultimate privacy-first physical workspace companion.
Log in or sign up for Devpost to join the conversation.