About

Bytes and Brains LLC explores and builds edge-deployed AI: open-source tools for running intelligence where the data lives.

The Vision

AI that runs at the edge, not in the cloud. Peers that are themselves experts, learning, specializing, and adapting to their local data distributions. Models that discover each other through the network, form fluid expert groups based on demand, and cache where they're needed. A system that works offline, adapts to the non-IID distributions each node actually sees, and routes intelligence to where it's needed. No central coordinator required.

The Journey

It started with gossip protocols, learning how distributed systems maintain themselves through local interactions. Along the way, we built product quantization and k-means clustering for vector compression. Tools that make high-dimensional peer-to-peer communication practical.

Each piece feeds into the next. The bytesandbrains Rust library ties them together.

Currently Working On

• BB-SIM: a generic simulator for decentralized overlay protocols, so every protocol in the library can be tested and measured end-to-end

What's Next

• Generalizing the core traits so any overlay protocol in the library plugs into the simulator and the training stack
• Integrating deep learning frameworks and methods to test federated, split, and gossip learning on top of the overlay

Founded by Matthew Love.
Contact: [email protected]