Roger Federated

A dichotomy in the AI transition is surfacing: Will we continue to delegate our decisional capacity to closed source models we cannot inspect nor steer and thus rely on far-fetched financial projections, or can we do better with tiny models running on consumer hardware? The answer is yes, we can do better. And that's where Roger Federated comes in. You and the community can now contribute to the next generation of AI. Not just an LLM with tools, but a purpose-trained agent that is inherently omni-modal. How? By locally finetuning a selected open-source foundation model on agentic rollout data, and subsequently aggregating the resulting encrypted model updates (not the data itself) securely with your selected federations.